Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Journal of Molecular Liquids 392 (2023) 123479

Contents lists available at ScienceDirect

Journal of Molecular Liquids


journal homepage: www.elsevier.com/locate/molliq

Investigation of the structure, stability, and relative solubility of psilocybin


in water and pure organic solvents: A molecular simulation study
Lucas Paul a , Cyril T. Namba-Nzanguim b,c , Aidani Telesphory d , Jehoshaphat Oppong Mensah e ,
Denis Mteremko f , Rene Costa g , Saidi Mohamedi Katundu h , Lucas P. Kwiyukwa i ,
Naserian Daniel Kambaine h,j , Julius Juvenary i , Sixberth Mlowe a , Geradius Deogratias i ,
Daniel M. Shadrack k , Andrew S. Paluch a,l,∗
a
Department of Chemistry, Dar es Salaam University College of Education, P.O. Box 2329, Dar es Salaam, Tanzania
b
Department of Chemistry, University of Buea, P.O. Box 63, Buea CM-00237, Buea, Cameroon
c
Center for Drug Discovery, Faculty of Science, University of Buea, P. O. Box 63, Buea, Cameroon
d
Department of Science and Laboratory Technology, Dar es Salaam Institute of Technology, P.O. Box 2958, Dar es salaam, Tanzania
e
Department of Chemistry, Kwame Nkrumah University of Science and Technology, PMB, Kumasi, Ghana
f
Department of Biological and Food Sciences, The Open University of Tanzania, Tanzania
g
Department of Physical and Environmental Sciences, The Open University of Tanzania, Tanzania
h
Department of Chemistry, College of Natural and Mathematical Sciences, The University of Dodoma, P.O. Box 338, Dodoma, Tanzania
i
Chemistry Department, College of Natural and Applied Sciences, University of Dar es Salaam, P.O. Box 35061, Dar es Salaam, Tanzania
j
Iphytos Company Tanzania LTD, P. O. Box 15014, Arusha, Tanzania
k
Chemistry Department, Faculty of Natural and Applied Sciences, St John’s University of Tanzania, Dodoma, Tanzania
l
Department of Chemical, Paper and Biomedical Engineering, Miami University, Oxford, OH 45056, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Psilocybin is an indole-based secondary metabolite found naturally in mushrooms which possesses several
Psilocybin pharmacological effects. Recently, a large number of experimental investigations have been conducted to
Structure characterize the pharmacology of psilocybin and its derivatives, and to develop synthetic pathways to
Stability
manufacture psilocybin. Nonetheless, current research on the physical characterization of psilocybin is limited in
Solubility
part due to legal restrictions. In the present study, we investigate two unique tautomers of psilocybin as depicted
Solvation
in the 2D chemical structure of psilocybin presented in the recent literature. Using a combination of electronic
structure calculations and molecular simulation, we are able to identify and characterize the thermodynamically
preferred tautomer. We additionally computed the solvation free energy and investigated the solvation structure
of psilocybin in water and 35 organic solvents. We find that hydrogen bonding between psilocybin and the
solvent dominates the solvation process. Considering the thermodynamically preferred tautomer, we find that
the solubility in water is greater than all of the studied organic solvents.

1. Introduction been found to greatly affect the mind, reduce anxiety and have hallu-
cinogenic activity [4]. Recently, psilocybin has been approved by the
Psilocybin (4-phosphoryloxy-N,N-dimethyltryptamine, Fig. 1), is an US Food and Drug Administration (FDA) for clinical studies for the
indole-based secondary metabolite found in “magic mushrooms.” Psilo- treatment of anxiety, depression and certain addictions [4]. Previous
cybin was first isolated by Albert Hofmann in 1957 and then syn- clinical trials have suggested that psilocybin is expected to have a great
thesized in 1958 [1]. It is produced by basidiomycetes species and impact on the treatments of psychiatric disorders.
extracted from several genera of fungi like the mushroom, Psilocybe Despite its known beneficial effects, psilocybin has clinical limita-
mexicana. tions related to (1) current research on the physical characterization
Psilocybin has shown a wide range of pharmacological applications of psilocybin are limited, and (2) neurotoxic activity [2]. Additionally,
including neuroprotective effects [2], mental disorders [3], and has psilocybin is a prodrug metabolized in vivo to psilocin by dephosphory-

* Corresponding author.
E-mail address: PaluchAS@MiamiOH.edu (A.S. Paluch).

https://doi.org/10.1016/j.molliq.2023.123479
Received 19 July 2023; Received in revised form 24 October 2023; Accepted 30 October 2023
Available online 7 November 2023
0167-7322/© 2023 Elsevier B.V. All rights reserved.
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Fig. 1. The chemical structure of psilocybin. The letter in parentheses is used


to indicate form “A” or “B”, labels introduced in this work to differentiate the
two tautomers. Form A is the common structure depicted in the literature, and
form B is the zwitterionic form reported as the product in ref. [7].

lation. To address this problem, it requires that psilocybin is dissolved


in an appropriate solvent and/or requires the use of nano-particles to
Fig. 2. A simple illustration representative of the solvation free energy calcula-
reduce the neurotoxic activity and enhance specific delivery. In addi-
tions performed in this work. The solvation free energy of the solute (compo-
tion to recent clinical investigations, recent successful efforts have been solv solv
nent 1) is computed in arbitrary solvent I and II, Δ𝐺1,I and Δ𝐺1,II , respectively,
made to develop synthetic routes to manufacture psilocybin [5–10]. which is related to the transfer free energy of the solute from solvent I to
Despite this, work to physically characterize psilocybin is limited, tran
II, Δ𝐺1,I→II tran
. In turn, Δ𝐺1,I→II is directly related to equilibrium concentration
in part due to legal restrictions. As a simple example, two different 2D (moles/volume or mass/volume) of the solute in solvent II and I, 𝑐1,II and 𝑐1,I ,
chemical structures of psilocybin have been used recently in the lit- respectively, where 𝑅 is the molar gas constant and 𝑇 is the absolute tempera-
erature on the development of synthetic routes for synthesis. See for ture.
example, ref. [9] versus [7]. In ref. [7], the 2D structure differs from
the rest of literature, wherein the authors report the product as zwit-
terionic psilocybin. Moreover, a knowledge and understanding of the
solubility of psilocybin in a range of solvents is lacking. Psilocybin is
known to be soluble in water, slightly soluble in methanol and acetone,
and insoluble in chloroform and hexane [3,11]. While knowledge of the
solubility is necessary to overcome clinical limitations in addition to
knowledge of aqueous solubility as it relates to bioavailability, pharma-
ceuticals in general need to satisfy various solubility criteria before they
may be manufactured at an industrial scale at a reasonable expense, and
then formulated for delivery [12–14]. With regards to production, the
synthesis is carried out in solution. Once synthesized, psilocybin must
be separated and purified. Alternatively, it may be necessary to selec-
tively extract psilocybin from its natural environment. Subsequently,
psilocybin needs to be formulated for delivery. If a tablet is to be made
to be administered orally, solvents will be necessary for the crystal-
lization process. Each of these processes can have different solubility
requirements, complicating design. In a typical manufacturing process, Fig. 3. A simple illustration representative of the free energy calculations per-
solvents account for 80–90% of the utilized material [12,14]. Given the formed to compare psilocybin form A and B. First, calculations are performed
large and complex chemistry of psilocybin, finding suitable solvents for to compute the Gibbs free energy of form A and B in the ideal gas phase, from
each stage of the manufacturing process using experimentation alone which we compute the Gibbs free energy of reaction for the interconversion of
rxn,ig
would be challenging. form A to B in an ideal gas, Δ𝐺A→B . This is combined with the solvation free
In the present study, we use a combination of electronic structure energy of form A and B in a common solvent, Δ𝐺Asolv and Δ𝐺Bsolv , respectively,
calculations and atomistic molecular simulation to determine and char- to obtain the Gibbs free energy of reaction for the interconversion of form A
rxn rxn
acterize the thermodynamically favored structure of psilocybin. We to B in solution, Δ𝐺A→B . In turn, Δ𝐺A→B is directly related to the equilibrium
additionally perform solvation free energy calculations for psilocybin concentration (moles/volume or mass/volume) of form B and A, 𝑐B and 𝑐A , re-
spectively, where 𝑅 is the molar gas constant and 𝑇 is the absolute temperature.
in water and 35 organic solvents to determine the relative solubility
and subsequently to probe the underlying solvation mechanism. The
use of molecular simulation is advantageous in that it is not only able molar concentration. In this way, Δ𝐺1solv is a direct measure of the
to quantitatively predict macroscopically observed properties of inter- solute-solvent intermolecular interactions. A negative value of Δ𝐺1solv
est, but can offer insight into the underlying molecular-level driving indicates that the solute prefers to be in solution (compared to the non-
forces [15–19]. interacting ideal gas state); the more negative the value of Δ𝐺1solv the
greater the affinity of the solute for the solution. When Δ𝐺1solv is com-
2. Methodology puted for the same solute in multiple solvents, the difference in Δ𝐺1solv
between two solvents may be related to the corresponding transfer free
In this work, we seek to understand the structure, stability, and rel- energy, which in turns can be related to the macroscopically observable
ative solubility of psilocybin (form A and B, see Fig. 1) in water and the relative concentration (mass or molar) of the solute at equilibrium (or
35 organic solvents which ref. [12] lists as the most commonly used in relative solubility) [20,21].
the chemical industry (see Tables 1 and 2). This is accomplished with In Fig. 2 we depict the case of a common solute (psilocybin form A
a combination of free energy calculations and detailed structural anal- or B) in multiple solvents. When comparing psilocybin forms A and B,
ysis. In all cases, we will have a single solute (psilocybin form A or both forms will have a different ideal gas reference state. The premise
B) molecule infinitely dilute in solution. The premise of the solvation of quantitatively comparing forms A and B are depicted in Fig. 3. Elec-
free energy calculations is depicted in Fig. 2. The solvation free energy, tronic structure calculations are performed to compute the Gibbs free
Δ𝐺1solv , corresponds to the change in Gibbs free energy upon taking the energy of forms A and B in the ideal gas phase (vacuum). From this,
solute from a non-interacting ideal gas phase to solution at the same we can compute the Gibbs free energy of reaction for the interconver-

2
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Table 1
A summary of the dimensionless solvation free energy of psilocybin form A in water and the
35 organic solvents which ref. [12] lists as the most commonly used in the chemical industry.
The solvents are classified by chemical functionality following ref. [12]. The subscript in the
reported solvation-free energy corresponds to the uncertainty in the three decimal places.
The solvation-free energy is broken down into the Coulombic (Coul, 𝑚 = 10 to 14) and LJ
(𝑚 = 0 to 10) contribution, where the total solvation-free energy corresponds to the sum of
these two contributions.

Δ𝐺1solv ∕ (𝑅𝑇 ) form A


Solvent Class CAS Coul LJ Total

ethanol alcohol 64-17-5 −25.350043 −18.808050 −44.158066


butanol alcohol 71-36-3 −23.627206 −20.638061 −44.265215
2-ethylhexanol alcohol 104-76-7 −22.674078 −22.163087 −44.837117
isobutanol alcohol 78-83-1 −21.796093 −20.784064 −42.580157
isopropanol alcohol 67-63-0 −24.168087 −20.158056 −44.326104
methanol alcohol 67-56-1 −26.714045 −16.313037 −43.027058
propanol alcohol 67-63-0 −23.134052 −20.058056 −43.192076
propylene glycol alcohol 57-55-6 −29.416057 −18.134090 −47.550107
acetone ketone 67-64-1 −22.795024 −19.630041 −42.425048
methyl ethyl ketone ketone 78-93-3 −21.539049 −20.608045 −42.147067
methyl isobutyl ketone ketone 108-10-1 −19.690048 −21.377078 −41.067092
methyl isopropyl ketone ketone 563-80-4 −21.299033 −21.121051 −42.420061
mesityl oxide ketone 141-79-7 −24.213046 −20.920055 −45.133072
ethylene bromide halogenated 106-93-4 −9.838026 −23.690065 −33.528070
chloroform halogenated 67-66-3 −12.272027 −23.389045 −35.660052
1,2-dichloroethane halogenated 107-06-2 −11.056022 −22.434046 −33.489051
dichloromethane halogenated 75-09-2 −13.179015 −22.080039 −35.259042
tetrachloroethylene halogenated 127-18-4 −0.582004 −24.652054 −25.234054
carbon tetrachloride halogenated 56-23-5 −0.368002 −24.342052 −24.728052
dimethylformamide amide 68-12-2 −31.142040 −19.276065 −50.418076
1,4-dioxane ether 123-91-1 −19.411031 −20.985077 −40.396083
butyl ether ether 142-92-1 −9.215028 −22.568061 −31.783067
ethyl ether ether 60-29-7 −12.440047 −21.689038 −34.128060
diisopropyl ether ether 108-20-3 −11.200038 −21.978061 −33.178072
tetrahydrofuran ether 109-99-9 −16.351064 −22.582053 −38.932083
tert-butyl methyl ether ether 1634-04-4 −12.748030 −21.858050 −34.607059
dimethyl sulfoxide sulfur 67-68-5 −33.576059 −17.353073 −50.930094
pyridine amine 110-86-1 −20.101035 −21.413049 −41.514060
acetonitrile nitrile 75-05-8 −18.675026 −20.027039 −38.702047
ethyl acetate esters 141-78-6 −22.748037 −20.724054 −43.471065
cyclohexane aliphatic 110-82-7 −0.170002 −23.339072 −23.509072
hexane aliphatic 110-54-3 −0.130001 −22.749042 −22.879042
toluene aromatic 108-88-3 −10.362031 −22.687051 −33.050059
xylene aromatic 1330-20-7 −10.333031 −23.371060 −33.704068
water water 7732-18-5 −42.718037 0.729091 −41.989098

rxn,ig
sion of form A to B in an ideal gas, Δ𝐺A→B . This can be combined with Simulations were performed for psilocybin (form A and B, see Fig. 1)
the solvation free energy of form A and B in a common solvent, to ob- in water and the 35 organic solvents which ref. [12] lists as the most
tain the Gibbs free energy of reaction for the interconversion of form commonly used in the chemical industry (see Tables 1 and 2). Psilo-
rxn . The Gibbs free energy of reaction in turn
A to B in solution, Δ𝐺A→B cybin form A corresponds to the common structure depicted in the
is related to the macroscopically observable relative concentration (or literature, and form B is the zwitterionic form reported as the product in
population) of form B relative to form A [20]. ref. [7]. Force fields were selected following the work of Caleman et al.
[24,25]. We first attempted to use the all-atom version of the optimized
3. Computational details potential for liquid simulations (OPLS-AA) [26–30], where the LigPar-
Gen web server was used to generate force field parameters for psilo-
3.1. Molecular simulation cybin [31,32]. However, bonded parameters were missing to model the
phosphate group, which led to instabilities in the simulations. As a re-
sult, psilocybin and the organic solvents were modeled with the General
All interactions were modeled using a “class I” potential energy
( ) AMBER Force Field version 2 (GAFF2) as implemented in the AMBER 20
function where all non-bonded intermolecular interactions 𝑈nb were
simulation suite [33–35] with AM1-BCC partial charges [36,37]. Param-
accounted for using a combined Lennard-Jones (LJ) plus fixed point
eters were generated using antechamber and converted from AMBER to
charge model of the form [22,23]
GROMACS format using ParmEd. Single molecule 3D structures of psilo-
[( ) ( )6 ]
𝜎𝑖𝑗 12 𝜎𝑖𝑗 cybin and the 35 organic compounds were generated with Open Babel
( ) 1 𝑞𝑖 𝑞𝑗
𝑈nb 𝑟𝑖𝑗 = 4𝜀𝑖𝑗 − + (1) 2.3.2 [38,39] by performing a systematic conformation search to iden-
𝑟𝑖𝑗 𝑟𝑖𝑗 4𝜋𝜀0 𝑟𝑖𝑗
tify the lowest energy conformer followed by geometry optimization,
where 𝑟𝑖𝑗 is the separation distance between sites 𝑖 and 𝑗, 𝜀𝑖𝑗 is the well- all using the General Amber Force Field (GAFF) [34] with Gasteiger
depth of the LJ potential, 𝜎𝑖𝑗 is the distance at which the LJ potential is partial charges [40]. Water was modeled using the rigid TIP4P/2005
zero, and 𝑞𝑖 and 𝑞𝑗 are the partial charges of sites 𝑖 and 𝑗, respectively. model [41].

3
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Table 2 All of the molecular dynamics simulations were performed using


A summary of the dimensionless solvation free energy of psilocybin form B in GROMACS 2020.2 [51–54]. For psilocybin and the organic solvents,
water and the 35 organic solvents which ref. [12] lists as the most commonly all bond lengths involving hydrogen were constrained using P-LINCS
used in the chemical industry. The solvents are classified by chemical function- [44,55,56]. Water was modelled as completely rigid using the SET-
ality following ref. [12]. The subscript in the reported solvation-free energy
TLE algorithm [57,58]. The Verlet neighbor list was used, and LJ in-
corresponds to the uncertainty in the three decimal places. The solvation free
teractions were cut-off at 1.4 nm with long-range analytic dispersion
energy is broken-down into the Coulombic (Coul, 𝑚 = 10 to 14) and LJ (𝑚 = 0 to
10) contribution, where the total solvation free energy corresponds to the sum
corrections applied to the energy and pressure [22,23,44,45]. Lorentz-
of these two contributions. Berthelot mixing rules were used for unlike LJ sites [22]. The elec-
trostatic terms were evaluated with the smooth particle-mesh-Ewald
Δ𝐺1solv ∕ (𝑅𝑇 ) form B
method (SPME) with tin-foil boundary conditions [44,45,59] with real
Solvent Class Coul LJ Total space interactions truncated at 1.4 nm. The SPME B-spline was order
ethanol alcohol −46.657133 −22.787044 −69.444140 4, the Fourier spacing was 0.12 nm, and the relative tolerance between
butanol alcohol −42.796187 −24.796059 −67.591196 long and short-range energies was 10−8 . The equations of motion were
2-ethylhexanol alcohol −38.394289 −26.578085 −64.972302
integrated with a timestep of 2 fs, the time constant for the thermostat
isobutanol alcohol −44.116061 −23.252063 −67.368124
isopropanol alcohol −44.147106 −23.915056 −68.062120 was 1 ps and the time constant for the barostat was 4 ps.
methanol alcohol −50.022126 −19.335035 −69.356131 All of the GROMACS force field files used in the present study along
propanol alcohol −43.908204 −24.120054 −68.028211 with sample GROMACS input files are provided in the Supporting In-
propylene glycol alcohol −54.710139 −22.722085 −77.432163 formation accompanying the electronic version of this manuscript.
acetone ketone −35.067053 −23.777040 −58.844067
methyl ethyl ketone ketone −32.279052 −24.702045 −56.980069
methyl isobutyl ketone ketone −29.405078 −25.694054 −55.098095 3.1.1. Free energy calculations
methyl isopropyl ketone ketone −29.996056 −25.367051 −55.363076 The solvation free energy, Δ𝐺1solv , for the solute infinitely dilutes in
mesityl oxide ketone −31.792069 −25.498054 −57.290088 water and the 35 organic solvents at 298.15 K and 1 bar was calcu-
ethylene bromide halogenated −20.256049 −29.021063 −49.277080 lated using a multi-stage free energy perturbation method [60–63] with
chloroform halogenated −26.138057 −27.886044 −54.025072 the multi-state Bennett’s acceptance ratio method (MBAR) [64–67] fol-
1,2-dichloroethane halogenated −22.768041 −26.832045 −49.600061
lowing our previous work [18,68–70]. A “soft-core” potential was used
dichloromethane halogenated −27.917031 −26.348038 −54.265049
tetrachloroethylene halogenated −0.989007 −29.626052 −30.615052 to couple/decouple the solute-solvent intermolecular LJ interactions.
carbon tetrachloride halogenated −0.701004 −29.104051 −29.805051 Stage (𝑚) dependent decoupling parameters, 𝜆LJ 𝑚 and 𝜆𝑚 , controlled
elec

dimethylformamide amide −42.772090 −24.510063 −67.281109 the LJ and electrostatic intermolecular interactions, respectively. The
1,4-dioxane ether −30.064055 −25.936076 −56.001094 decoupling parameters varied from 0 to 1. When 𝜆LJ 𝑚 = 𝜆𝑚 = 1, the so-
elec

butyl ether ether −14.288062 −26.896054 −41.184082 lute is fully coupled to the system. When 𝜆𝑚 = 𝜆𝑚 = 0, the solute is
LJ elec

ethyl ether ether −19.920032 −25.417036 −45.337050 decoupled from the system. The use of a soft-core potential is impor-
diisopropyl ether ether −17.226052 −26.031044 −43.257068
tant in that while it gives the correct limiting value of the LJ potential,
tetrahydrofuran ether −26.136037 −27.123051 −53.260062
tert-butyl methyl ether ether −20.115044 −25.934047 −46.049064
it additionally allows nearly decoupled molecules to overlap with fi-
nite energy (and hence finite probability), thereby increasing the phase
dimethyl sulfoxide sulfur −46.081137 −22.763069 −68.844153
space overlap between neighboring states. The soft-core potential had
pyridine amine −33.589054 −26.086050 −59.675074
the form [71–73]
acetonitrile nitrile −34.442033 −24.388037 −58.830049
( )
ethyl acetate esters −32.742155 −24.970052 −57.712164 𝑈LJ
sc
𝑟𝑖𝑗 ; 𝑚
cyclohexane aliphatic −0.399002 −27.948065 −28.348065
hexane aliphatic −0.340002 −26.782041 −27.122041 ⎧ ⎫
⎪ 𝜎𝑖𝑗12 𝜎𝑖𝑗6 ⎪
toluene aromatic −19.776046 −27.187048 −46.963066 = 4𝜆LJ
𝑚 𝜀𝑖𝑗 ⎨ [ ] − [ ( ) ] ⎬ (2)
( ) 2
xylene aromatic −19.026066 −27.741057 −46.767087 ⎪ 1 − 𝜆𝐿𝐽 𝛼 𝜎 6 + 𝑟6 1 − 𝜆𝐿𝐽
𝑚 𝛼𝐿𝐽 𝜎𝑖𝑗 + 𝑟𝑖𝑗 ⎪
6 6
⎩ 𝑚 𝐿𝐽 𝑖𝑗 𝑖𝑗 ⎭
water water −76.021347 −4.117084 −80.138357
where 𝛼LJ is a constant, which had a value of 1/2. The electrostatic term
in the intermolecular potential was decoupled linearly as
While either Monte Carlo or molecular dynamics could be used to ( ) 1 𝑞𝑖 𝑞𝑗
compute configurational properties, here we use molecular dynamics. 𝑈elec 𝑟𝑖𝑗 ; 𝑚 = 𝜆elec
𝑚 4𝜋𝜀 𝑟 (3)
0 𝑖𝑗
All of the simulations consisted of a single solute (psilocybin form A
or B) molecule infinitely dilute in solution. The number of solvent At each stage 𝑚, an independent MD simulation was performed. For
molecules was chosen to obtain a cubic box with an edge length of each stage 𝑚 the total simulation time was 22 ns with the first 2 ns dis-
approximately 4.6 nm at 298.15 K and 1 bar. Initial structures were carded from analysis as equilibration. The change in the Hamiltonian
generated using Packmol [42,43]. This was followed by 3,000 steepest with the current configuration between stage 𝑚 and the other stages is
descent minimization steps to remove any bad contacts that might have computed every 0.20 ps. This is saved for subsequent post-simulation
resulted from the packing. We next performed two steps of equilibra- analysis with MBAR [67] to determine Δ𝐺1solv . This analysis was per-
tion in an 𝑁𝑃 𝑇 ensemble at 298.15 K and 1 bar with the equations of formed using the Python implementation of MBAR (PyMBAR) and the
motion integrated using the Verlet leap-frog algorithm [22,23,44,45]. GROMACS analysis script distributed with it [74]. The analysis script
First, we performed a 2 ns equilibration using the Berendsen thermo- has implemented an autocorrelation analysis so that only uncorrelated
stat and barostat [44–46]. This was followed by 12 ns of equilibration samples are used to determine Δ𝐺1solv and the corresponding uncertainty
using the stochastic velocity rescaling thermostat [44,47–49] and the [75–77].
Parrinello-Rahman barostat [50]. The final structure from this series A total of 15 different stages were used for the free energy calcula-
of simulations was then used as the initial structure for our free en- tions where 𝑚 = 0 corresponds to a non-interacting (ideal gas) state and
ergy calculations, as will be described momentarily. The final 𝑁𝑃 𝑇 𝑚 = 14 is a fully interacting system. From 𝑚 = 1 to 10 the LJ interactions
equilibration for psilocybin form A and B in ethanol, propylene glycol, were increased from 𝜆LJ 𝑚 = 0.1 to 1.0 in 10 equal increments of 0.1. Elec-
acetone, chloroform, carbon tetrachloride, pyridine, toluene, cyclohex- trostatic interactions were increased in a square root fashion following
ane, and water was continued for an additional 100 ns which was used 𝜆elec
𝑚 ={0.50, 0.71, 0.87, 1.00} from 𝑚 = 11 to 14. The change in free en-
for subsequent structural analysis. ergy in going from 𝑚 = 0 to 10 (non-interacting to full LJ) we refer to

4
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

as the LJ contribution, Δ𝐺1solv,LJ , and the change in free energy in go- which are strong hydrogen bond acceptors. This could possibly suggest
ing from 𝑚 = 10 (full LJ with no electrostatic) to 14 (full LJ and full that psilocybin form A is primarily acting as a hydrogen bond donor,
electrostatic) we refer to as the Coulombic contribution, Δ𝐺1solv,Coul . The or that there is a competition between interactions wherein psilocybin
total solvation-free energy is the sum of these two contributions. Put dif- form A acts as a hydrogen bond donor and acceptor leading to no net
ferently, the LJ contribution corresponds to the change in free energy difference in the ketones and alcohols. We will clarify this point mo-
in going from a state with no solute-solvent intermolecular interactions mentarily with detailed structural analysis.
(𝑚 = 0) to a state with solute-solvent LJ intermolecular interactions (but The most negative value of the Coulombic contribution to Δ𝐺1solv is
no electrostatic interactions, 𝑚 = 10). Within the state 𝑚 = 0, the solute for water. On the other hand, the LJ contribution to Δ𝐺1solv in water
may be found anywhere in the system with equal probability. It fol- is the only case where Δ𝐺1solv is positive, indicative of the difficulty
lows that an interesting feature of the LJ contribution is that it captures disrupting the hydrogen bonding network of water to accommodate
the change in free energy resulting from the solute cavity formation the solute. A positive value of the LJ contribution to Δ𝐺1solv indicates
in solution. The Coulombic contribution corresponds to the change in that psilocybin form A would prefer to be in a non-interacting ideal
free energy in going from a state with solute-solvent LJ intermolec- gas state than in water with full LJ interactions (but no intermolecular
ular interactions but no electrostatic interactions (𝑚 = 10) to a state electrostatic interactions). This again underscores the important role of
with full solute-solvent LJ and electrostatic intermolecular interactions. hydrogen bonding in the solvation of psilocybin.
The Coulombic contribution will therefore capture the effect of solute- The computed values of Δ𝐺1solv along with the LJ and Coulombic
solvent intermolecular hydrogen bonding. contribution for psilocybin form B in water and the 35 organic solvents
The simulation parameters for the free energy calculations were the are summarized in Table 2. We again find that hydrogen bonding plays
same as the last step of equilibration except the equations of motion an important role. For the aliphatic alkanes and the non-hydrogen con-
were integrated with the GROMACS “stochastic dynamics” integrator, taining halogens, the Coulombic contribution to Δ𝐺1solv is again close to
corresponding to stochastic or velocity Langevin dynamics integrated 0. However, we find that the effect of hydrogen bonding is noticeably
with the leap-frog algorithm [44,45,78]. The time constant for the more important with psilocybin form B as compared to form A. Going
stochastic thermostat was 1.0 ps. This change is necessary as a local from chloroform to carbon tetrachloride, with form B the Coulombic
thermostat is required to correctly control the temperature of the de- contribution and the total Δ𝐺1solv increase by approximately 25𝑅𝑇 and
coupled and weakly coupled solute molecule. 24𝑅𝑇 , respectively, which is more than double that observed for form
Sample GROMACS input files are provided in the Supporting Infor- A.
mation accompanying the electronic version of this manuscript. Comparing the alcohols and ketones for form B, the Coulombic con-
tribution and the total Δ𝐺1solv for the alcohols is now more negative
3.2. Electronic structure calculations than the ketones; this could suggest that psilocybin form B is a stronger
hydrogen bond acceptor than donor, leading to more favorable interac-
Electronic structure calculations were performed to determine the tions in the alcohols as compared to the ketones. Moreover, the most
Gibbs free energy of forms A and B in the ideal gas phase. The ini- negative value of the Coulombic contribution and the total Δ𝐺1solv both
tial 3D molecular structures of forms A and B were generated using now correspond to water. Interestingly, for water, the LJ contribution
the Avogadro package [79]. The geometries were brought to the min- to Δ𝐺1solv is now also negative.
imum conformational energy using the Universal Force Field (UFF) Fig. 4 is a parity plot of the Coulombic and LJ contribution, and the
[80] available within the software. In each case, a geometry optimiza- total Δ𝐺1solv for psilocybin form B versus form A. In all cases, the contri-
tion and frequency calculation was subsequently performed with the butions and total Δ𝐺1solv for form B are less than form A. This indicates
B3LYP/cc-PVTZ level of theory/basis set using Gaussian 16 [81,82]; a greater stabilization of form B in the solution as compared to form
the equilibrium geometries were confirmed by the absence of imagi- A. Considering the Coulombic contribution, we find that the values for
nary frequencies. From the frequency calculation the ideal gas-phase forms B and A are well correlated with an 𝑅2 of 0.93. Moreover, the
Gibbs free energy was computed as the “Sum of electronic and thermal line of best fit has a slope of 1.62. The change in the Coulombic con-
free energies” [83]. tribution is systematic; the more negative the Coulombic contribution
for form A, the greater the decrease for form B. This corroborates the
4. Results and discussion finding that hydrogen bonding is important in the solvation of psilo-
cybin, with the effect greatest for form B. The Coulombic contribution
4.1. Solvation free energy calculations decreases on average by 12.3𝑅𝑇 , with a standard deviation of 7.4𝑅𝑇 .
The largest decrease is for water (33.3𝑅𝑇 ), and the smallest decrease is
The computed values of Δ𝐺1solv along with the LJ and Coulombic for hexane (0.2𝑅𝑇 ). Considering the LJ contribution, we find that the
contribution for psilocybin form A in water and the 35 organic solvents values for forms B and A are very well correlated with an 𝑅2 of 0.98.
are summarized in Table 1. Based on the structure of psilocybin, we ex- For this case, the slope of the line of best fit is 1. The LJ contribution
pect hydrogen bonding to be important in the solvation process. This is decreases on average by 4.3𝑅𝑇 with a standard deviation of 0.6𝑅𝑇 .
reflected in our results where for the aliphatic alkanes (cyclohexane and
hexane) and the non-hydrogen containing halogens (tetrachloroethy- 4.2. Free energy of form A and B
lene and carbon tetrachloride) the Coulombic contribution to Δ𝐺1solv is
close to 0, and as a result, the value of the total Δ𝐺1solv is greater (or less Using electronic structure calculations, the Gibbs free energy of form
negative) than all of the others. Moreover, consider the comparison of B relative to form A in the ideal gas phase is –8.53𝑅𝑇 , which corre-
chloroform and carbon tetrachloride. The two molecules differ in that sponds to the Gibbs free energy of reaction for the interconversion of
chloroform contains a single hydrogen which is a weak hydrogen bond rxn,ig
forms A to B, Δ𝐺A→B . The negative value indicates that form B is the
donor [84,85]. When the hydrogen is replaced with chlorine (to give
preferred form. Form B corresponds to the zwitterionic form of psilocy-
carbon tetrachloride), the Coulombic contribution and the total Δ𝐺1solv rxn,ig
bin reported in ref. [7]. The magnitude is significant, where Δ𝐺A→B is
increase by approximately 12𝑅𝑇 and 11𝑅𝑇 , respectively.
related to the macroscopically observable relative molar (or mass) con-
Comparing the alcohols and ketones, we find that the Coulombic ig ig
centration of form B (𝑐B ) relative to form A (𝑐A ) in the ideal gas-phase
contribution and the total Δ𝐺1solv values are comparable. Whereas al-
( rxn,ig )
cohols are capable of both donating and accepting a hydrogen bond, ig
𝑐B Δ𝐺A→B
ketones are acceptors only. The most negative values of Δ𝐺1solv are = exp − = 5.08 × 103 (4)
ig
𝑐A 𝑅𝑇
in dimethylformamide (DMF) and dimethyl sulfoxide (DMSO), both of

5
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Table 3
A summary of the dimensionless solvation free energy of psilocybin
form B relative to A, ΔΔ𝐺1solv ∕ (𝑅𝑇 ), and the dimensionless Gibbs free
energy of reaction for the interconversion of form A to B in solution,
rxn
Δ𝐺A→B ∕ (𝑅𝑇 ). The subscript in the reported values corresponds to the
uncertainty in the two decimal places.

A→B
Solvent Class ΔΔ𝐺1solv ∕ (𝑅𝑇 ) rxn
Δ𝐺A→B ∕ (𝑅𝑇 )
ethanol alcohol −25.2915 −33.8215
butanol alcohol −23.3329 −31.8629
2-ethylhexanol alcohol −20.1432 −28.6732
isobutanol alcohol −24.7920 −33.3220
isopropanol alcohol −23.7416 −32.2716
methanol alcohol −26.3314 −34.8614
propanol alcohol −24.8422 −33.3722
propylene glycol alcohol −29.8819 −38.4219
acetone ketone −16.4208 −24.9508
methyl ethyl ketone ketone −14.8310 −23.3710
methyl isobutyl ketone ketone −14.0313 −22.5613
methyl isopropyl ketone ketone −12.9410 −21.4810
mesityl oxide ketone −12.1611 −20.6911
ethylene bromide halogenated −15.7511 −24.2811
chloroform halogenated −18.3709 −26.9009
1,2-dichloroethane halogenated −16.1108 −24.6408
dichloromethane halogenated −19.0106 −27.5406
tetrachloroethylene halogenated −5.3807 −13.9107
carbon tetrachloride halogenated −5.0807 −13.6107
dimethylformamide amide −16.8613 −25.4013
1,4-dioxane ether −15.6113 −24.1413
butyl ether ether −9.4011 −17.9311
ethyl ether ether −11.2108 −19.7408
diisopropyl ether ether −10.0810 −18.6110
tetrahydrofuran ether −14.3310 −22.8610
tert-butyl methyl ether ether −11.440 9 −19.980 9
dimethyl sulfoxide sulfur −17.9118 −26.451 8
Fig. 4. A parity plot of the Coulombic (Coul, 𝑚 = 10 to 14) and LJ (𝑚 = 0 to 10)
contribution, and the total solvation free energy for psilocybin form B versus pyridine amine −18.1610 −26.6910
form A. (See Tables 1 and 2.) The solid line corresponds to 𝑦 = 𝑥 and is drawn acetonitrile nitrile −20.1307 −28.6607
as a reference, and 𝑚 and 𝑅2 correspond to the slope and squared correlation ethyl acetate esters −14.2418 −22.7718
coefficient, respectively, of the line of best fit to the data.
cyclohexane aliphatic −4.8410 −13.3710
hexane aliphatic −4.2406 −12.7806
rxn,ig toluene aromatic −13.9109 −22.4509
The value of Δ𝐺A→B can be combined with Δ𝐺1solv for psilocybin
xylene aromatic −13.0611 −21.6011
form A and B to determine the Gibbs free energy of reaction for the
rxn (see Fig. 3). A sum- water water −38.1537 −46.6837
interconversion of form A to B in solution, Δ𝐺A→B
mary of the computed values of Δ𝐺A→B rxn , along with Δ𝐺solv of form B
1
relative to form A, ΔΔ𝐺1solv , is provided in Table 3. In all cases, ΔΔ𝐺1solv precipitation [86]. This was a point made in ref. [7] wherein a synthetic
is negative as already discussed, corresponding to a greater stabilization process to was proposed to produce the zwitterionic form of psilocybin
of form B relative to A in solution. Since Δ𝐺A→B rxn = Δ𝐺rxn,ig + ΔΔ𝐺solv ,
A→B 1 (form B). The trend is consistent with the literature that psilocybin is
rxn is more negative than Δ𝐺 rxn,ig
we have that Δ𝐺A→B A→B
. This indicates that soluble in water, slightly soluble in methanol and acetone, and insol-
form B is thermodynamically preferred over form A, where in all cases uble in chloroform and hexane [3,11]. Interestingly, for form A (see
the preference is significant. Table 1), water is not the solvent in which psilocybin has the great-
Having identified form B as the thermodynamically preferred form est solubility. For form A, the solubility is greatest is the polar aprotic
of psilocybin, we revisit Table 2 for the Δ𝐺1solv of form B in the stud- solvents dimethyl sulfoxide and dimethylformamide.
ied solvents. As shown in Fig. 2, the relative value of Δ𝐺1solv is related
to the macroscopically observable relative molar (or mass) concentra- 4.3. Structural analysis
tion in solution at equilibrium. The value of Δ𝐺1solv is most negative
in water, indicating the psilocybin has the greatest solubility in wa- Psilocybin form A and B (see Fig. 1) are capable of forming inter-
ter. The next closest solvent is propylene glycol, for which the Δ𝐺1solv and intra-molecular hydrogen bonds, and are able to act as both a
differs by 2.706𝑅𝑇 , corresponding to a molar (or mass) concentration hydrogen bond donor and acceptor. The free energy calculations sug-
approximately 15 times greater in water than in propylene glycol. The gested that form B is a stronger hydrogen bond acceptor than donor,
difference is even greater in the common solvent ethanol, the next clos- leading to more favorable interactions in alcohols as compared to ke-
est solvent, where the relative concentration is over 44-thousand times tones. On the other hand, for form A, there was no major difference
greater in water than in ethanol. Based on these findings, the use of between solvation in the ketones and alcohols, suggesting that either
anti-solvent crystallization may prove a promising technique to pu- form A is a stronger hydrogen bond donor or that there is a competi-
rify psilocybin from aqueous solutions. Anti-solvent crystallization is tion between interactions wherein psilocybin form A acts as a hydrogen
a technique to recover the solute with a relatively high solubility from bond donor and acceptor leading to no net difference. To clarify these
solution, which avoids temperature control as in cooling crystallization. points, structural analysis was performed for psilocybin form A and B
Instead, a concentrated solution is mixed with another miscible solvent in ethanol, propylene glycol, acetone, chloroform, carbon tetrachloride,
in which the solubility is much lower, leading to supersaturation and pyridine, toluene, cyclohexane, and water to better understand the sol-

6
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

lar interactions with the (unprotonated) tertiary amine N (N1). On the


other hand, psilocybin form B exhibits stronger intramolecular interac-
tion between the equivalent O atoms (O3 and O4) with the H atom in
the ammonium (protonated tertiary amine, H2), in addition to a weaker
intramolecular interaction between O1 and H2. As discussed earlier, us-
ing electronic structure calculations we found that form B had a lower
free energy as compared to form A in the ideal gas state. From the anal-
ysis here in carbon tetrachloride, form B may be favored over A due to
the presence of stronger intramolecular interactions.
The CMat for psilocybin in acetone is found in Fig. 7 C and D for
form A and B, respectively. Acetone is a polar aprotic solvent that is
only capable of accepting hydrogen bonds. Psilocybin form A has two
phosphate hydrogen atoms (H1 and H2) which strongly, and pyrrole
H attached to the ring N (H3) which weakly interacts (intermolecular)
with the acetone O (O1). Form A also exhibits the same intramolecu-
lar interactions as found in carbon tetrachloride, only weaker. On the
other hand, for form B we observe the same strong intramolecular inter-
actions as in carbon tetrachloride. Additionally, weaker intermolecular
interactions exist between the acetone O atom (O1) and the phosphate
H atom (H1), and both H atoms bonded to an N atom (H2 and H3) with
different intensities.
The CMat for psilocybin in pyridine is found in Fig. 7 E and F for
form A and B, respectively. Similar to acetone, pyridine is a polar apro-
tic solvent. As a result, we find that for both forms the interactions are
analogous to the case of acetone. This is in support of the computed
values of Δ𝐺1solv discussed earlier (see Tables 1 and 2), wherein the nu-
merical values for acetone and pyridine did not differ significantly.
Next, the CMat for psilocybin in ethanol is found in Fig. 7 G and
H for form A and B, respectively. Ethanol is a polar protic solvent ca-
pable of both donating and accepting hydrogen bonds. As a result, we
observe both inter- and intra-molecular hydrogen bonding, where psilo-
cybin both donates and accepts hydrogen bonds. In form A, we observe
the same intermolecular interactions wherein psilocybin acts the hydro-
gen bond donor in addition to the same intramolecular interactions as
observed in acetone. However, we additionally observe hydrogen bond-
Fig. 5. Structural images for psilocybin form A and B with the atomic labels
ing between the ethanol H atom (H1) and the O atoms in the phospate
used during structural analysis [87,88].
group of psilocybin (O1, O2, O3, and O4), with the strongest interac-
tion with O4. Likewise for form B, we observe the same intermolecular
vation mechanism. Structural images for psilocybin and the solvents are interactions wherein psilocybin acts the hydrogen bond donor in ad-
provided in Figs. 5 and 6, respectively, with the atomic labels used dur- dition to the same intramolecular interactions, as observed in acetone.
ing analysis and discussion here. However, we also observe hydrogen bonding between the ethanol H
atom (H1) and the O atoms in the phospate group of psilocybin (O1,
4.3.1. Connection matrix analysis (CMat) O2, O3, and O4). We also find that with the presence of these addi-
First, to gain an overview the inter- (solute-solvent) and intra- tional intermolecular hydrogen bonds, the intensity of the intramolec-
molecular (solute-solute) hydrogen bonding, we performed a connec- ular interactions decreases. Additionally, for form A the strength of the
tion matrix (CMat) analysis [87,88]. In Fig. 7 we show results for the interactions wherein psilocybin was a donor (H1,2–O1) and acceptor
CMat analysis for the solvents capable for participating in hydrogen (O4–H1,2) were comparable. On the other hand, for form B psilocybin
bonding (ethanol, propylene glycol, acetone, pyridine, and water) and was a stronger acceptor (O3,4–H1) than donor (H1–O1), with an in-
as a counter example, carbon tetrachloride. Generally, the connection tensity greater than that exhibited by form A. This corroborates with
matrix is made up of two parts; the left part shows the interaction con- the result of the free energy calculations, suggesting that form B is a
nection between the promising proton donor and acceptor while the stronger hydrogen bond acceptor than donor, resulting in a lower sol-
right side is the heat map which displays a measure of the interaction vation free energy in the alcohols as compared to the ketones. This also
strength. The interaction strength is quantified using the first peak of provides clarity in that form A participates in hydrogen bonding as both
the local density (i.e., the un-normalized radial distribution function), an acceptor and donor with comparable strength. The competition be-
where the greater the intensity (or density) and the shorter the dis- tween acting as a hydrogen bond donor and acceptor must therefore be
tance, the stronger the interaction. A matrix box filled with a black the reason of the only minor difference in the Columbic contribution to
cross suggests there are no interactions within a cutoff distance of 350 the solvation free energy observed in the ketones and alcohols.
pm. Within the analysis 1–2, 1–3, and 1–4 intra-molecular interactions The CMat for psilocybin in water is found in Fig. 7 I and J for form A
are also excluded, as are solvent-solvent intermolecular interactions. and B, respectively. Also similar to ethanol, water is a polar protic sol-
Within Fig. 7 we place potential hydrogen bond accepting sites on the vent, although we expect that water is a stronger hydrogen bond donor
rows, and potential hydrogen bond donating sites on the columns. and acceptor. We find that for both forms the interactions analogous
The CMat for psilocybin in carbon tetrachloride is found in Fig. 7 to the case of ethanol. As compared to ethanol, we do find that the
A and B for form A and B, respectively. As a solvent carbon tetrachlo- intensities (i.e., local densities) of the intermolecular hydrogen bonds
ride is not capable of participating in hydrogen bonding. Both form A are higher, confirming stronger hydrogen bonds between psilocybin
and B display the existence of only intramolecular hydrogen bonding. and water as compared to ethanol. This is confirmed by the computed
For form A, the phosphate H atoms (H1 and H2) show intramolecu- values of Δ𝐺1solv discussed earlier. We found that the Coulombic con-

7
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Fig. 6. Structural images for the select solvents with the atomic labels used during structural analysis [87,88].

tribution was much more negative in water as compared to ethanol. with the CMat analysis. We note that the RDF intensity followed the
Moreover, the LJ contribution in water was larger (less favorable) in trend ethanol > propyl glycerol > pyridine > acetone > water > carbon
part as a consequence of the necessity to break strong water-water in- tetrachloride. We note that while in the CMat analysis we observed a
termolecular hydrogen bonds in order to create a cavity for the solute relatively large local density of water around psilocybin, the intensity of
in solution. Moreover, the Coulombic contribution for form B in water the RDF is noticeably relatively smaller. Quantiatively this results from
was more negative (i.e., more favorable) than A confirming that form B the larger bulk density of water compared to the other solvents, and
is a stronger hydrogen bond acceptor. Additionally, while for form A in is indicative that in water psilocybin is competing with strong water-
ethanol the strength of the interactions wherein psilocybin was a donor water intermolecular interactions. As noted earlier, this is reflected in
(H1,2–O1) and acceptor (O4–H1,2) were comparable, in water the case the results of the solvation free energy calculations.
where psilocybin is an acceptor is stronger. For form B, the same trend To confirm the observation of hydrogen bonding between psilocybin
is observed in water as in ethanol wherein psilocybin appears to be a and the solvents, in Figure S3 of the SI we show the results of a com-
stronger hydrogen bond acceptor. If we consider the hydrogen bond bined distribution function (CDF) analysis. For the CDF, we plot the
donating and accepting scale used by the solubility parameter based results of an angle distribution function (ADF) analysis versus a RDF
method MOSCED (Modified Separation of Cohesive Energy Density), analysis, to show the angle versus distance for the interactions of inter-
the hydrogen bond donating and accepting strength of ethanol are com- est, with the scale indicative of the frequency of occurrence. A hydrogen
parable (12.58 and 13.29, respectively). On the other hand, water is bond is characterized by a short interaction distance (taken here to be
a much stronger hydrogen bond donor (52.78) than acceptor (15.86) less than 450 pm), and specific interaction angle (taken here to be be-
[89–91]. The result of water being a much stronger hydrogen bond tween 130◦ and 180◦ or between 0◦ and 50◦ , depending on how the
donor is likely why the intermolecular interaction wherein form A is angle is measured) [87,88]. The results of the CDF analysis confirm the
acting as the acceptor in water is stronger than the case where form A presence of hydrogen bonding.
is acting as a donor. Last, in Figure S4 of the SI we show the results of a spatial distribu-
Finally, the CMat for psilocybin in propylene glycol is found in Fig. 7 tion function (SDF) analysis. Results are presented for psilocybin forms
K and L for form A and B, respectively. Similar to ethanol, propylene A and B in ethanol and carbon tetrachloride. The solvents were selected
glycol is a polar protic solvent. However, propylene glycol contains two following the RDF analysis as the solvents with the largest and smallest
hydroxyl groups that is able to form both inter- and intra-molecular hy- RDF intensities, respectively. The SDF if used to visualize the 3D sol-
drogen bonds [91–96]. We find that for both forms the interactions are vation of psilocybin by the solvent, where the iso-surfaces correspond
analogous to the case of ethanol. We note that while the CMat shows a to local densities of a given value. The SDF analysis was visualized us-
lower intensity for the intermolecular hydrogen bonds between psilocy- ing the Visual Molecular Dynamics (VMD) software. For the case of
bin and propylene glycol (or local density maximum), this is the result ethanol we observe clustering of the solvent around the polar regions
of psilocybin interacting with two hydroxyl groups. of psilocybin. On the other hand, for carbon tetrachloride, we observe
a propensity toward the non-polar regions of psilocybin.
4.4. Additional structural analysis
4.5. Electronic structure calculations
Additional structural analysis was performed in support of CMat,
with the results presented in the supporting information (SI) accompa- 4.5.1. Stability and vibrational spectra
nying the electronic version of this manuscript. First, results from the The stability or possible existence of the two forms of psilocybin (A
radial distribution (RDF) are presented in Figures S1 and S2 of the SI, and B) can be further characterized from the analysis of the optimized
for form A and B, respectively, and in Table S1. As compared to CMat, geometries and vibrational frequencies computed using electronic struc-
the RDF corresponds to the local relative to bulk density as a func- ture calculations. The electronic structure calculation (i.e., DFT) results
tion of distance. The results from the RDF analysis are in agreement showed that form B is relatively more stable than form A with a –21.15

8
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

Fig. 8. Comparison of the computed IR spectra of psilocybin form A and B.

Table 4
Vibrational modes of forms A and B calculated at the B3LYP/cc-PVTZ level of
theory in the gaseous phase. The symbols 𝜔 = wagging vibration, 𝛿 = bending
vibration, 𝜈 = stretching vibration and 𝜌 = rocking vibration (in plane).
A (cm−1 ) B (cm−1 ) Experimental (cm−1 ) Tentative assignment of vibrations
465.31 465.18 461.25 𝜔PO4
754.12 734.33 746.77 𝛿CCC(indole) + 𝜈NR2 + 𝜔CH(indole)
780.67 798.43 785.93 𝜌CH2
854.54 857.97 855.93 𝛿CCC(indole) +𝜈P−OH
938.44 891.25 920.69 𝜈P−OH
1050.94 1034.53 1042.33 𝜌OH
1102.01 1102.99 1101.24 𝜌CH(indole) + 𝜌CH2
1242.35 1221.98 1229.69 𝛿CCC + 𝜈C=N(indole)
1336.70 1287.23 1348.55 𝜈P=O+𝜌CH2
1446.72 1465.36 1442.19 𝜈C=C + 𝜈C=N(indole)
− 1823.24 − 𝜈N−H⋯O
3189.58 3184.56 3179.96 𝜈CH(benzene)

Each form belongs to the C1 symmetry point group consisting of


36 atoms which accounts for 102 vibrational degrees of freedom. Con-
former A has been experimentally reported among the two conformers
and the IR spectral data are available for comparison [97]. Therefore,
in Table 4, we compare the calculated IR active vibrational modes for
form A with experimental IR frequencies and provide tentative spec-
tral assignments. Form A exhibits five bands with distinct intensities at:
170−540 cm−1 , 750−950 cm−1 , 1000−1660 cm−1 , 2900−3250 cm−1
and 3670−3830 cm−1 . It is worth noting that the calculated IR spectrum
for form A matches well with the experimentally reported spectra. For
form B, we observed bands with varying intensities at: 200-560 cm−1 ,
Fig. 7. The connection matrix (CMat) analysis of psilocybin form A and B in 720−900 cm−1 , 960−1100 cm−1 , 1200−1680 cm−1 . Conformer B ex-
carbon tetrachloride (CCl4 , A and B), acetone (C and D), pyridine (E and F), hibits a distinct peak at 1823.24 cm−1 originating from the abstraction
ethanol (G and H), water (I and J), and propylene glycol (K and L). The first of a hydrogen atom from the phosphate group to the dimethylamino
column of figures corresponds to psilocybin form A, and the second column group. Other bands observed include 3050−3200 cm−1 and 3650−3830
corresponds to psilocybin form B. cm−1 . In the region below 2000 cm−1 form B exhibits higher intensity
than those of A and vice versa for vibrations above 3000 cm−1 .
The IR spectra of both forms revealed distinct stretching vibrations
kJ/mol free energy difference between them. The absence of imaginary 𝜈, including the 𝜈N−H originating from the five-membered ring at 3677
frequencies confirms that the obtained geometries during optimization cm−1 . Also, the 𝜈O−H for the two terminal OH groups in form A were ob-
correspond to the minima on the potential energy surface. In order to served at 3817 cm−1 and 3826 cm−1 , while that of form B was detected
understand the differences in the two forms of psilocybin, their respec- at 3830 cm−1 . The out of plane 𝜈O−H exhibits a higher intensity than the
tive harmonic vibrational frequencies were analyzed, and the calculated one in-plane. The 𝜈C−H of the dimethylamine group were observed at
infrared (IR) spectra for both forms are presented in Fig. 8. The two around 2900 cm−1 and 3042–3049 cm−1 for A and B, respectively. The
forms are very similar, only that B is obtained when hydrogen is ab- CH2 units exhibit 𝜈C−H at 3027(3042) cm−1 for A and 3089(3116) cm−1
stracted from the acidic moiety (phosphate group) to the basic moiety for B. For form B, H is positioned between N and O, and one can observe
(dimethylamino group) resulting in a zwitterionic product also known that form B exhibits a sharp peak with the highest intensity at 1823.24
as an inner salt. cm−1 corresponding to N-H stretching with an average bond distance of

9
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

be relatively reactive when compared to form B. Based on this analysis,


it is recommended that further research should focus on characterizing
both forms of psilocybin. Investigating the chemical kinetic and ther-
modynamic properties of these two forms of psilocybin can shed light
on their respective chemical stability.

5. Summary and conclusion

Psilocybin is the active psychedelic compound known for hallucino-


genic properties present in “magic mushrooms.” Since its structural
elucidation and chemical synthesis in 1958, both synthetic and puri-
fied psilocybin have been subjected to clinical trials, with the findings
showing that the compound presents huge potential for the treatment of
Fig. 9. The HOMO and LUMO charge density distribution, HOMO/LUMO ener-
various mental health disorders. However, as a result of legal measures
gies and energy gap of psilocybin A and B. (Note that 1 eV = 96.485 kJ/mol
psilocybin use has been restricted to controlled medical and scientific
[100].)
research in most of the world. Its clinical applications are hampered by
poor solubility in most organic solvents, and the ability to study and
investigate the physical properties of psilocybin are limited.
In this work we investigate two different tautomers of psilocybin (la-
beled here as form A and B). While form A corresponds to the tautomer
commonly depicted in the literature, form B corresponds to the zwitte-
rionic form reported recently in ref. [7]. We expect that the presence
of two different chemical structures in the literature results from the
limited ability to experimentally investigate the physical properties of
psilocybin. We use molecular simulation to investigate the interaction
and solvation of psilocybin in a wide range of solvents. We additionally
used electronic structure calculations (DFT) to reveal that form B is the
thermodynamically favored tautomer. The calculated vibrational char-
acteristics in this study accord well with experimentally observed data.
Both exhibit two absorption bands with the higher energy band at 196
nm and 198 nm while the lower energy bands were observed at 252
and 250 nm for A and B, respectively.
The propensity of psilocybin to solvate in solvents was deter-
mined by using molecular simulation solvation free energy calculations,
Δ𝐺1solv , which is related to the macroscopically observable relative
Fig. 10. The electronic absorption spectra of psilocybin form A and B.
molar (or mass) concentration at equilibrium (or relative solubility)
[21]. The computed Δ𝐺1solv along with LJ and Coulombic contribu-
1.130 Å and a 1.422 Å O-H bond. Other characteristic vibrations and
tion showed that form B had a higher stabilization in solvents than
assignments are summarized in Table 4.
form A. Hydrogen bonding has a strong influence in the solvation pro-
The results of the electronic structure calculation reveal differences
cess. We found that form B is a stronger hydrogen bond acceptor than
in the geometrical parameters between the two forms. For instance, the
donor, leading to more favorable interactions in alcohols as compared
P-O bond lengths in the H2 PO4 group (form A) are 1.605, 1.467, 1.612
to ketones. On the other hand, for form A, there was no major dif-
and 1.594 Å, while those in HPO4 (form B) are 1.652, 1.485, 1.616
ference between solvation in the ketones and alcohols. Form A acts
and 1.529 Å. As can be observed hydrogen abstraction resulted in the
as a hydrogen bond donor and acceptor with comparable strength,
shortening of the P-O− bond and an increase in other P-O bonds.
leading to a competition between interactions in solution. This solva-
4.5.2. Optical and electronic characteristics tion mechanism was further confirmed using detailed structural analy-
Frontier molecular orbitals are widely used for the analysis of sta- sis.
bility and chemical reactivity of molecules [98,99]. Fig. 9 shows the Having identified form B as the thermodynamically preferred tau-
calculated highest occupied molecular orbital (HOMO), lowest unoccu- tomer, the results of the solvation free energy showed that the solubility
pied molecular orbital (LUMO) energies, energy gaps and HOMO and in water is greater than in any of the studied organic solvents. This could
LUMO charge density distribution. The electronic absorption spectra for present opportunities for anti-solvent crystallization to purify psilocybin
psilocybin A and B are presented in Fig. 10. Both HOMO and LUMO of from an aqueous solution as suggested in ref. [7] wherein a synthetic
form B are up-shifted by 0.29 eV and 0.32 eV, respectively (Note that 1 process to was proposed to produce the zwitterionic form of psilocybin
eV = 96.485 kJ/mol [100]). The two forms depict comparable HOMO (form B). If one were interesting in improving the solubility relative to
and LUMO charge distributions. The simulated electronic spectra show water, future studies might consider the use of alternative solvents such
that both forms exhibit similar electronic transitions characterized by as ionic liquids or deep eutectic solvents. The present study suggests
two bands in the ultraviolet region of the electromagnetic spectrum. that in order to increase the solubility relative to water, one would re-
The high energy bands were observed around 196–198 nm resulting quire a solvent that is a stronger hydrogen bond donor than water. The
from 𝑛 → 𝜋 ∗ transitions and the lower energy bands were observed at authors acknowledge that the accuracy of DFT predictions can be influ-
252 nm and 250 nm for A and B, respectively; the lower energy bands enced by the choice of the level of theory. In this study, the reported
are observed due to 𝜋 → 𝜋 ∗ transitions. At absorption maximum wave- results are specifically based on a comparison with experimental data
lengths, form B’s electronic spectra are slightly blue-shifted compared for form A, revealing a standard deviation of 5.14 cm−1 . To improve
to form A; consequently leading to an increased energy gap for B, as DFT predictions, it is recommended to evaluate different DFT function-
seen in Fig. 10. The smaller energy gap for A indicates that form A can als; especially those specialized in modeling hydrogen bonding.

10
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

CRediT authorship contribution statement [8] J. Fricke, F. Blei, D. Hoffmeister, Enzymatic synthesis of psilocybin, Angew. Chem.
Int. Ed. 56 (2017) 12352–12355.
[9] J. Fricke, C. Lenz, J. Wick, F. Blei, D. Hoffmeister, Production options for psilocy-
Lucas Paul: Conceptualization, Formal analysis, Investigation,
bin: making of the magic, Chem. Eur. J. 25 (2019) 897–903.
Methodology, Project administration, Visualization, Writing – original [10] A.M. Sherwood, P. Meisenheimer, G. Tarpley, R.B. Kargbo, An improved, practical,
draft, Writing – review & editing. Cyril T. Namba-Nzanguim: For- and scalable five-step synthesis of psilocybin, Synthesis 52 (2020) 688–694.
mal analysis, Investigation, Writing – original draft, Writing – review [11] Scientific Working Group for the Analysis of Seized Drugs (SWGDRUG) Mono-
& editing. Aidani Telesphory: Formal analysis, Investigation, Writing graph: Psilocybin, https://www.swgdrug.org/Monographs/PSILOCIN.pdf. (Ac-
cessed 26 June 2019).
– review & editing. Jehoshaphat Oppong Mensah: Formal analysis, [12] K. Grodowska, A. Parczewski, Organic solvents in the pharmaceutical industry,
Investigation, Writing – review & editing. Denis Mteremko: Formal Acta Pol. Pharm. 67 (2010) 3–12.
analysis, Investigation, Writing – review & editing. Rene Costa: Formal [13] D.J.C. Constable, C. Jimenez-Gonzalez, R.K. Hendersen, Perspective on solvent use
analysis, Investigation, Writing – review & editing. Saidi Mohamedi in the pharmaceutical industry, Org. Process Res. Dev. 11 (2007) 133–137.
[14] R. Liu (Ed.), Water-Insoluble Drug Formulation, 2nd ed., CRC Press, Boca Raton,
Katundu: Formal analysis, Investigation, Writing – review & editing.
FL, 2008.
Lucas P. Kwiyukwa: Formal analysis, Investigation, Writing – review [15] C.A. Slabber, C.D. Grimmer, R.S. Robinson, Solution conformations of curcumin in
& editing. Naserian Daniel Kambaine: Formal analysis, Investiga- DMSO, J. Nat. Prod. 79 (2016) 2726–2730.
tion, Writing – review & editing. Julius Juvenary: Formal analysis, [16] J. Ilnytskyi, T. Patsahan, O. Pizio, On the properties of the curcumin molecule in
Investigation, Writing – review & editing. Sixberth Mlowe: Formal water. Exploration of the OPLS-united atom model by molecular dynamics com-
puter simulation, J. Mol. Liq. 223 (2016) 707–715.
analysis, Investigation, Writing – review & editing. Geradius Deogra-
[17] N.D. Kambaine, D.M. Shadrack, S.A.H. Vuai, Conformations and stability of cap-
tias: Conceptualization, Formal analysis, Investigation, Methodology, saicin in bulk solvents: a molecular dynamics study, J. Mol. Liq. 345 (2022)
Visualization, Writing – original draft, Writing – review & editing. 117794.
Daniel M. Shadrack: Conceptualization, Formal analysis, Investiga- [18] H.C. Da Silva, A.S. Paluch, L.T. Costa, W.B. De Almeida, Thermodynamic and struc-
tion, Methodology, Visualization, Writing – original draft, Writing – tural description of relative solubility of the flavonoid rutin by DFT calculations
and molecular dynamics simulations, J. Mol. Liq. 341 (2021) 117214.
review & editing. Andrew S. Paluch: Conceptualization, Formal anal- [19] L. Paul, G. Deogratias, D.M. Shadrack, C.N. Mudogo, K.M. Mtei, R.L. Machunda,
ysis, Investigation, Methodology, Project administration, Resources, A.S. Paluch, R. Ntie-Kang, A molecular investigation of the solvent influence on
Visualization, Writing – original draft, Writing – review & editing. inter- and intra-molecular hydrogen bond interaction of linamarin, Processes 10
(2022) 352.
[20] S.A. Rodriguez, J.V. Tran, S.J. Sabatino, A.S. Paluch, Predicting octanol/water par-
Declaration of competing interest
tition coefficients and pKa for the SAMPL7 challenge using the SM12, SM8 and
SMD solvation models, J. Comput.-Aided Mater. Des. 36 (2022) 687–705.
The authors declare that they have no known competing financial [21] To be clear, we are unable to predict solubility. We are unable to predict the “ab-
interests or personal relationships that could have appeared to influence solute” (not relative) solubility as we would need to know the fugacity of the solid
relative to the fugacity of pure liquid psilocybin (typically referred to as the ideal
the work reported in this paper.
solubility). As illustrated in Figures 2 and 3, we can relate the solvation free ener-
gies to what would appear to be a relative solubility. But this is not the equilibrium
Data availability solubility but would be equivalent to the equilibrium partitioning (concentration)
of an infinitely dilute solute between two solvents (i.e., octanol/water partition
Data will be made available on request. coefficient). We can draw the equivalence to solubility if the solute is assumed
dilute.
[22] A.R. Leach, Molecular Modelling: Principles and Applications, 2nd ed., Pearson
Acknowledgements Education Limited, Harlow, England, 2001.
[23] D. Frenkel, B. Smit, Understanding Molecular Simulation: From Algorithms to Ap-
All of the molecular dynamics simulations were performed at the plications, 2nd ed., Academic Press, San Diego, CA, 2002.
[24] C. Caleman, P.J. van Maaren, M. Hong, J.S. Hub, L.T. Costa, D. van der Spoel, Force
Ohio Supercomputer Center on the Pitzer Cluster [101]. Resources were
field benchmark of organic liquids: density, enthalpy of vaporization, heat capaci-
graciously provided to support a workshop on free energy calculations ties, surface tension, isothermal compressibility, volumetric expansion coefficient,
by Andrew S. Paluch delivered at the Dar es Salaam University College and dielectric constant, J. Chem. Theory Comput. 8 (2012) 61–74.
of Education (DUCE), Tanzania. [25] virtualchemistry.org, http://virtualchemistry.org/. (Accessed 11 May 2022).
[26] W.L. Jorgensen, D.S. Maxwell, J. Tirado-Rives, Development and testing of the
OPLS all-atom force field on conformational energetics and properties of organic
Appendix A. Supplementary material liquids, J. Am. Chem. Soc. 118 (2014) 11225–11236.
[27] J.Z. Vilseck, J. Tirado-Rives, W.L. Jorgensen, Evaluation of CM5 charges for
Supplementary material related to this article can be found online condensed-phase modeling, J. Chem. Theory Comput. 10 (2014) 2802–2812.
at https://doi.org/10.1016/j.molliq.2023.123479. [28] M.J. Robertson, J. Tirado-Rives, W.L. Jorgensen, Improved peptide and protein
torsional energetics with the OPLS-AA force field, J. Chem. Theory Comput. 11
(2015) 3499–3509.
References [29] L.S. Dodda, J.Z. Vilseck, J. Tirado-Rives, W.L. Jorgensen, 1.14*CM1A-LBCC: lo-
calized bond-charge corrected CM1A charges for condensed-phase simulations, J.
[1] T. Passie, J. Seifert, U. Schneider, H.M. Emrich, The pharmacology of psilocybin, Chem. Theory Comput. 121 (2017) 3864–3870.
Addict. Biol. 7 (2002) 357–364. [30] M.J. Robertson, Y. Qian, M.C. Robinson, J. Tirado-Rives, W.L. Jorgensen, Develop-
[2] V.K. Chaturvedi, S. Agarwal, K.K. Gupta, P.W. Ramteke, M.P. Singh, Medicinal ment and testing of the OPLS-AA/M force field for RNA, J. Chem. Theory Comput.
mushroom: boon for therapeutic applications, 3 Biotech 8 (2018) 1–20. 15 (2019) 2734–2742.
[3] M. Coppola, F. Bevione, R. Mondola, Psilocybin for treating psychiatric disorders: a [31] L.S. Dodda, I. Cabeza de Vaca, J. Tirado-Rives, W.L. Jorgensen, LigParGen web
psychonaut legend or a promising therapeutic perspective?, J. Xenobiot. 12 (2022) server: an automatic OPLS-AA parameter generator for organic ligands, Nucleic
41–52. Acids Res. 45 (2017) W331–W336.
[4] D.E. Nichols, Psilocybin: from ancient magic to modern medicine, J. Antibiot. 73 [32] LigParGen, http://zarbi.chem.yale.edu/ligpargen/. (Accessed 11 May 2022).
(2020) 679–686. [33] D.A. Case, K. Belfon, I.Y. Ben-Shalom, S.R. Brozell, D.S. Cerutti, T.E. Cheatham III,
[5] A.M. Adams, N.A. Kaplan, Z. Wei, J.D. Brinton, C.S. Monnier, A.L. Enacopol, T.A. V.W.D. Cruzeiro, T.A. Darden, R.E. Duke, G. Giambasu, M.K. Gilson, H. Gohlke,
Ramelot, J.A. Jones, In vivo production of psilocybin in E. coli, Metab. Eng. 57 A.W. Goetz, R. Harris, S. Izadi, S.A. Izmailov, K. Kasavajhala, A. Kovalenko, R.
(2019) 111–119. Krasny, T. Kurtzman, T.S. Lee, S. LeGrand, P. Li, C. Lin, J. Liu, T. Luchko, R. Luo,
[6] N. Milne, P. Thomsen, N. Mølgaard Knudsen, P. Rubaszka, M. Kristensen, I. Borod- V. Man, K.M. Merz, Y. Miao, O. Mikhailovskii, G. Monard, H. Nguyen, A. Onufriev,
ina, Metabolic engineering of Saccharomyces cerevisiae for the de novo production F. Pan, S. Pantano, R. Qi, D.R. Roe, A. Roitberg, C. Sagui, S. Schott-Verdugo, J.
of psilocybin and related tryptamine derivatives, Metab. Eng. 60 (2020) 25–36. Shen, C. Simmerling, N.R. Skrynnikov, J. Smith, J. Swails, R.C. Walker, J. Wang,
[7] R.B. Kargbo, A. Sherwood, A. Walker, N.V. Cozzi, R.E. Dagger, J. Sable, K. O’Hern, L. Wilson, R.M. Wolf, X. Wu, Y. Xiong, Y. Xue, D.M. York, P. Kollman, AMBER 20
K. Kaylo, T. Patterson, G. Tarpley, P. Meisenheimer, Direct phosphorylation of (2020).
psilocin enables optimized cGMP kilogram-scale manufacture of psilocybin, ACS [34] J. Wang, R.M. Wolf, J.W. Caldwell, P.A. Kollman, D.A. Case, Development and
Omega 5 (2020) 16959–16966. testing of a general amber force field, J. Comput. Chem. 25 (2004) 1157–1174.

11
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

[35] J. Wang, W. Wang, P.A. Kollman, D.A. Case, Automatic atom type and bond type [68] G.B. Fuerst, R.T. Ley, A.S. Paluch, Calculating the fugacity of pure, low volatile
perception in molecular mechanical calculations, J. Mol. Graph. Model. 25 (2006) liquids via molecular simulation with application to acetanilide, acetaminophen,
247–260. and phenacetin, Ind. Eng. Chem. Res. 54 (2015) 9027–9037.
[36] A. Jakalian, B.L. Bush, D.B. Jack, C.I. Bayly, Fast, efficient generation of high- [69] R.T. Ley, G.B. Fuerst, B.N. Redeker, A.S. Paluch, Developing a predictive form
quality atomic charges. AM1-BCC model: I. Method, J. Comput. Chem. 21 (2000) of MOSCED for nonelectrolyte solids using molecular simulation: application to
132–146. acetanilide, acetaminophen, and phenacetin, Ind. Eng. Chem. Res. 55 (2016)
[37] A. Jakalian, D.B. Jack, C.I. Bayly, Fast, efficient generation of high-quality atomic 5415–5430.
charges. AM1-BCC model: II. Parameterization and validation, J. Comput. Chem. [70] S.J. Sabatino, A.S. Paluch, Predicting octanol/water partition coefficients using
23 (2002) 1623–1641. molecular simulation for the SAMPL7 challenge: comparing the use of neat and
[38] N.M. O’Boyle, M. Banck, C.A. James, C. Morley, T. Vandermeersch, G.R. Hutchin- water saturated 1-octanol, J. Comput.-Aided Mater. Des. 35 (2021) 1009–1024.
son, Open babel: an open chemical toolbox, J. Cheminform. 3 (2011) 33. [71] T.C. Beutler, A.E. Mark, R.C. van Schaik, P.R. Gerber, W.F. van Gunsteren, Avoid-
[39] Open Babel: the Open Source Chemistry Toolbox, http://openbabel.org/wiki/ ing singularities and numerical instabilities in free energy calculations based on
Main_Page. (Accessed 26 June 2019). molecular simulations, Chem. Phys. Lett. 222 (1994) 529–539.
[40] J. Gasteiger, M. Marsili, A new model for calculating atomic charges in molecules, [72] M.R. Shirts, V.S. Pande, Solvation free energies of amino acid side chain analogs for
Tetrahedron Lett. 34 (1978) 3181–3184. common molecular mechanics water models, J. Chem. Phys. 122 (2005) 134508.
[41] J.L.F. Abascal, C. Vega, A general purpose model for the condensed phase of water: [73] T. Steinbrecher, D.L. Mobley, D.A. Case, Nonlinear scaling schemes for Lennard-
TIP4P/2005, J. Chem. Phys. 123 (2005) 234505. Jones interactions in free energy calculations, J. Chem. Phys. 127 (2007) 214108.
[42] L. Martínez, R. Andrade, E.G. Birgin, J.M. Martínez, Packmol: a package for build- [74] PyMBAR: Python implementation of the multistate Bennett acceptance ratio
ing initial configurations for molecular dynamics simulations, J. Comput. Chem. (mbar), https://github.com/choderalab/pymbar. (Accessed 20 May 2021).
30 (2009) 2157–2164. [75] J.D. Chodera, W.C. Swope, J.W. Pitera, C. Seok, K.A. Dill, Use of the weighted
[43] Packmol: packing optimization for molecular dynamics simulations, http://www. histogram analysis method for the analysis of simulated and parallel tempering
ime.unicamp.br/~martinez/packmol/. (Accessed 31 March 2021). simulations, J. Chem. Theory Comput. 3 (2007) 26–41.
[44] GROMACS development team, GROMACS documentation: release 2020.2, https:// [76] P.V. Klimovich, M.R. Shirts, D.L. Mobley, Guidelines for analysis of free energy
manual.gromacs.org/documentation/2020.2/manual-2020.2.pdf, 2020. calculations, J. Comput.-Aided Mol. Des. 29 (2015) 397–411.
[45] H.J.C. Berendsen, Simulating the Physical World: Hierarchial Modeling from Quan- [77] J.D. Chodera, A simple method for automated equilibration detection in molecular
tum Mechanics to Fluid Dynamics, Cambridge University Press, New York, NY, simulations, J. Chem. Theory Comput. 12 (2016) 1799–1805.
2007. [78] W.F. van Gunsteren, H.J.C. Berendsen, A leap-frog algorithm for stochastic dynam-
[46] H.J.C. Berendsen, J.P.M. Postma, A. DiNola, J.R. Haak, Molecular dynamics with ics, Mol. Simul. 1 (1988) 173–185.
coupling to an external bath, J. Chem. Phys. 81 (1984) 3684–3690. [79] M.D. Hanwell, D.E. Curtis, D.C. Lonie, T. Vandermeersch, E. Zurek, G.R. Hutchi-
[47] G. Bussi, D. Donadio, M. Parrinello, Canonical sampling through velocity-rescaling, son, Avogadro: an advanced semantic chemical editor, visualization, and analysis
J. Chem. Phys. 126 (2007) 014101. platform, J. Cheminform. 4 (2012) 1–17.
[48] G. Bussi, M. Parrinello, Stochastic thermostats: comparison of local and global [80] A.K. Rappe, C.J. Casewit, K.S. Colwell, W.A. Goddard III, W.M. Skiff, Uff, a full
schemes, Comput. Phys. Commun. 179 (2008) 26–29. periodic table force field for molecular mechanics and molecular dynamics simu-
[49] G. Bussi, T. Zykova-Timan, M. Parrinello, Isothermal-isobaric molecular dynamics lations, J. Am. Chem. Soc. 114 (1992) 10024–10039.
using stochastic velocity rescaling, J. Chem. Phys. 130 (2009) 074101. [81] C.J. Cramer, Essentials of Computational Chemistry, John Wiley & Sons Ltd, Chich-
[50] M. Parrinello, A. Rahman, Polymorphic transitions in single crystals: a new molec- ester, West Sussex, England, 2002.
ular dynamics method, J. Appl. Phys. 52 (1981) 7182–7190. [82] M.J. Frisch, G.W. Trucks, H.B. Schlegel, G.E. Scuseria, M.A. Robb, J.R. Cheese-
[51] B. Hess, C. Kutzner, D. van der Spoel, E. Lindal, GROMACS 4: algorithms for highly man, G. Scalmani, V. Barone, G.A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A.V.
efficient, load-balanced, and scalable molecular simulation, J. Chem. Theory Com- Marenich, J. Bloino, B.G. Janesko, R. Gomperts, B. Mennucci, H.P. Hratchian, J.V.
put. 4 (2008) 435–447. Ortiz, A.F. Izmaylov, J.L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini,
[52] S. Pronk, S. Páll, R. Schulz, P. Larsson, P. Bjelkmar, R. Apostolov, M.R. Shirts, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V.G. Za-
J.C. Smith, P.M. Kasson, D. van der Spoel, B. Hess, E. Lindahl, GROMACS 4.5: krzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R.
a high-throughput and highly parallel open source molecular simulation toolkit, Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T.
Bioinformatics 29 (2013) 845–854. Vreven, K. Throssell, J.A. Montgomery Jr., J.E. Peralta, F. Ogliaro, M.J. Bearpark,
[53] M.J. Abraham, T. Murtola, R. Schulz, S. Páll, J.C. Smith, B. Hess, E. Lindahl, GRO- J.J. Heyd, E.N. Brothers, K.N. Kudin, V.N. Staroverov, T.A. Keith, R. Kobayashi, J.
MACS: high performance molecular simulations through multi-level parallelism Normand, K. Raghavachari, A.P. Rendell, J.C. Burant, S.S. Iyengar, J. Tomasi, M.
from laptops to supercomputers, SoftwareX 1–2 (2015) 19–25. Cossi, J.M. Millam, M. Klene, C. Adamo, R. Cammi, J.W. Ochterski, R.L. Martin, K.
[54] GROMACS: Fast, flexible, free, http://www.gromacs.org/. (Accessed 31 March Morokuma, O. Farkas, J.B. Foresman, D.J. Fox, Gaussian 16, Revision C.01, 2019.
2021). [83] Thermochemistry in Gaussian, https://gaussian.com/thermo/. (Accessed 25 May
[55] B. Hess, H. Bekker, H.J.C. Berendsen, J.G.E.M. Fraaije, LINCS: a linear constraint 2022).
solver for molecular simulations, J. Comput. Chem. 18 (1997) 1463–1472. [84] J.J. Shepard, S.K. Callear, S. Imberti, J.S.O. Evans, C.G. Salzmann, Microstruc-
[56] B. Hess, P-LINCS: a parallel linear constraint solver for molecular simulation, J. tures of negative and positive azeotropes, Phys. Chem. Chem. Phys. 18 (2016)
Chem. Theory Comput. 4 (2008) 116–122. 19227–19235.
[57] J.P. Ryckaert, G. Ciccotti, H.J.C. Berendsen, Numerical integration of the Cartesian [85] G.A. Jeffrey, An Introduction to Hydrogen Bonding, Oxford University Press, Inc.,
equations of motion of a system with constraints: molecular dynamics of n-alkanes, New York, NY, 1997.
J. Comput. Phys. 23 (1977) 327–341. [86] M. Shahid, G. Sanxaridou, S. Ottoboni, L. Lue, C. Price, Exploring the role of anti-
[58] S. Miyamoto, P.A. Kollman, SETTLE: an analytical version of the SHAKE and RAT- solvent effects during washing on active pharmaceutical ingredient purity, Org.
TLE algorithms for rigid water models, J. Comput. Chem. 13 (1992) 952–962. Process Res. Dev. 25 (2021) 969–981.
[59] M. Deserno, C. Holm, How to mesh up Ewald sums. I. A theoretical and numer- [87] M. Brehm, B. Kirchner, TRAVIS - a free analyzer and visualizer for Monte Carlo
ical comparison of various particle mesh routines, J. Chem. Phys. 109 (1998) and molecular dynamics trajectories, J. Chem. Inf. Model. 51 (2011) 2007–2023.
7678–7693. [88] M. Brehm, M. Thomas, S. Gehrke, B. Kirchner, TRAVIS - a free analyzer for trajec-
[60] K.S. Shing, S.T. Chung, Computer simulation methods for the calculation of solu- tories from molecular simulation, J. Chem. Phys. 152 (2020) 164105.
bility in supercritical extraction systems, J. Phys. Chem. 91 (1987) 1674–1681. [89] M.J. Lazzaroni, D. Bush, C.A. Eckert, T.C. Frank, S. Gupta, J.D. Olson, Revision of
[61] D.A. Kofke, P.T. Cummings, Quantitative comparison and optimization of meth- MOSCED parameters and extension to solid solubility calculations, Ind. Eng. Chem.
ods for evaluating the chemical potential by molecular simulation, Mol. Phys. 92 Res. 44 (2005) 4075–4083.
(1997) 973–996. [90] P. Dhakal, S.N. Roese, E.M. Stalcup, A.S. Paluch, Application of MOSCED to predict
[62] D.A. Kofke, P.T. Cummings, Precision and accuracy of staged free-energy perturba- hydration free energies, Henry’s constants, octanol/water partition coefficients,
tion methods for computing the chemical potential by molecular simulation, Fluid and isobaric azeotropic vapor-liquid equilibrium, J. Chem. Eng. Data 63 (2018)
Phase Equilib. 150–151 (1998) 41–49. 352–364.
[63] C. Chipot, A. Pohorille (Eds.), Free Energy Calculations: Theory and Applications [91] P. Dhakal, S.N. Roese, M.A. Lucas, A.S. Paluch, Predicting limiting activity co-
in Chemistry and Biology, Springer Series in Chemical Physics, vol. 86, Springer, efficients and phase behavior from molecular structure: expanding MOSCED to
New York, NY, 2007. alkanediols using group contribution methods and electronic structure calcula-
[64] C.H. Bennett, Efficient estimation of free energy differences from Monte Carlo data, tions, J. Chem. Eng. Data 63 (2018) 2586–2598.
J. Comp. Physiol. 22 (1976) 245–268. [92] T.M. Usacheva, V.I. Zhuravlev, N.V. Lifanova, V.K. Matveev, Molecular dynamics
[65] M.R. Shirts, E. Bair, G. Hooker, V.S. Pande, Equilibrium free energies from nonequi- models and thermodynamic characteristics of hydrogen bonds in 1, 2-ethanediol,
librium measurements using maximum-likelihood methods, Phys. Rev. Lett. 91 Russ. J. Phys. Chem. A 91 (2017) 1056–1063.
(2003) 140601. [93] A. Jindal, S. Vasudevan, Conformation of ethylene glycol in the liquid state: intra-
[66] N. Lu, J.K. Singh, D.A. Kofke, Appropriate methods to combine forward and reverse versus intermolecular interactions, J. Phys. Chem. B 121 (2017) 5595–5600.
free-energy perturbation averages, J. Chem. Phys. 118 (2003) 2977–2984. [94] D.L. Crittenden, K.C. Thompson, M.J.T. Jordan, On the extent of intramolecular
[67] M.R. Shirts, J.D. Chodera, Statistically optimal analysis of samples from multiple hydrogen bonding in the gas-phase and hyrdrated 1, 2-ethanediol, J. Phys. Chem.
equilibrium states, J. Chem. Phys. 129 (2008) 124105. A 109 (2005) 2971–2977.

12
L. Paul, C.T. Namba-Nzanguim, A. Telesphory et al. Journal of Molecular Liquids 392 (2023) 123479

[95] O. Guvench, A.D. MacKerell Jr., Quantum mechanical analysis of 1, 2-ethanediol and anchoring groups: a dft/td-dft investigation, J. Mol. Graph. Model. 94 (2020)
conformational energetics and hydrogen bonding, J. Phys. Chem. A 110 (2006) 107480.
9934–9939. [99] S. Subashchandrabose, H. Saleem, Y. Erdogdu, G. Rajarajan, V. Thanikacha-
[96] C.J. Cramer, D.G. Truhlar, Quantum chemical conformational analysis of 1, 2- lam, Ft-Raman, ft-ir spectra and total energy distribution of 3-pentyl-2, 6-
ethanediol: correlation and solvation effects on the tendency to form internal diphenylpiperidin-4-one: dft method, Spectrochim. Acta, Part A, Mol. Biomol.
hydrogen bonds in the gase phase and in aqueous solution, J. Am. Chem. Soc. Spectrosc. 82 (2011) 260–269.
116 (1994) 3892–3900. [100] The NIST Reference on Constants, Units, and Uncertainty, https://physics.nist.gov/
[97] C.S. Esteves, E.M. de Redrojo, J.L.G. Manjón, G. Moreno, F.E. Antunes, G. Mon- cuu/Constants/. (Accessed 26 June 2023).
talvo, F.E. Ortega-Ojeda, Combining ftir-atr and opls-da methods for magic mush- [101] Ohio Supercomputer Center, Ohio Supercomputer Center, http://osc.edu/ark:
rooms discrimination, Forensic Chemistry 29 (2022) 100421. /19495/f5s1ph73, 1987.
[98] G. Deogratias, N. Seriani, T. Pogrebnaya, A. Pogrebnoi, Tuning optoelectronic
properties of triphenylamine based dyes through variation of pi-conjugated units

13

You might also like