African swine fever virus (ASFV) is a virus that belongs to the Asfarviridae family and is endemic in most

sub-Saharan African countries, where it infects warthogs and bush pigs and is often transmitted via soft
ticks. ASFV infections result in mild disease in warthogs and bush pigs but are much more serious in
domestic swine and wild boars, causing high mortality rates of between 90-100%. ASFV infections are
responsible for increasing swine mortality in several parts of the world, and there is no vaccine or
effective antiviral drug available to stop the epidemic. ASFV has spread to wild boar populations in
Europe and Asia, and the virus has a linear dsDNA genome of 170-195 kb with >150 open reading
frames. ASFV virulence is believed to be affected by multigene families, making it challenging to create
an effective vaccine. Currently, the p72 gene is the most frequently used gene for PCR diagnosis of ASFV,
and there are 24 ASFV genotypes described based on p72 sequences.

The study retrieved 48 complete genomes of African swine fever virus (ASFV) from GenBank, with a size
range of 170,000 to 200,000 base pairs. One of the genomes was excluded because it was identical to
another genome. The researchers recorded information about the country, date, original host (tick,
warthog, wild boar, or domestic pig), and virulence of the virus from the GenBank entries and original
literature. A summary of the 47 genomes used in the analysis is provided in Supplementary T.

The article describes the methods used to identify Pfam domains in ASFV genomes and custom profile
HMMs for MGFs. The hmmsearch function of HMMER-3.2.1 was used to identify the Pfam domains
encoded by ASFV genomes, while custom profile HMMs were prepared for MGF protein coding
sequences. The article also discusses the analysis of the ASFV encoded UK protein coding sequence and
the link between domain bit-scores and protein identity. The article provides details of the tools and
software used for the analysis and how to access them.

The study identified 82 domains that were present at least once per genome in 47 ASFV genomes using a
domain_i-Evalue cutoff of 0.0001. Seventeen domains were found multiple times in some genomes. The
domain content and scores were used to examine patterns in the ASFV genomes by generating a total
score for each domain and normalizing the values. Domains that showed >0.03 variance in their score
were retained and used for hierarchical clustering.

The article discusses a domain classification approach to study the African swine fever virus (ASFV)
genome. The UK gene's ORF encoding a 96 aa protein expressed early in ASFV infection was examined.
The study retrieved ASFV UK coding regions from GenBank, and an alignment of the proteins set was
shown. The study used the HMMER-3 search of the UK ORFs to provide a quantitative score (bit-score) of
the distance of the query domain from the model domain. The study identified all profile HMM domains
from the Pfam collection encoded in a set of 47 ASFV genomes and ordered them by hierarchical
clustering based on the Pfam + MGF domain scores. The approach was validated by clustering genomes
in nearly the same pattern as the p72 ML tree topology. The study identified diversity in the MGF
domains and eleven domains across different genotypes, which were not revealed from a p72 ML tree.

The article discusses the clustering pattern of MGF ORFs based on their custom profile HMMs, revealing
greater domain/functional variety in some genes than previously appreciated. The article also highlights
changes in MGF counts with ASFV genotype and between attenuated and virulent strains. The authors
suggest that changes in MGF numbers might result in altered viral properties, as seen in various strains.
The absence or presence of specific domains, such as Ank-4, is also noted in different genotypes.
The article discusses the use of genome-scale domain comparison method to examine paired ASFV
strains with reported differences in virulence. The authors highlight the importance of such analyses in
understanding the molecular basis for attenuation or virulence and guiding vaccine design efforts. An
example is given of a naturally occurring ASFV variant from Estonia that displayed attenuation in animal
tests, which was compared to contemporary viruses from Georgia using the domain classification tool to
identify changes in protein domains.

The article discusses a new method for characterizing ASFV-encoded protein diversity based on profile
HMM descriptions of conserved protein domains. The method allows for the identification of greater
diversity in the MGFs than previously noted, including the presence of unconventional MGFs that appear
distinct to specific strains of ASFV. The method also allows for the rapid assessment of qualitative
features of encoded domains and reports a bit-score for each identified domain, as well as copy number
changes in domains. The approach is useful for quality control of newly assembled genomes and for
exploring novel ASFV genomes as they are sequenced and annotated, as well as for comparing genomes
with varied clinical, epidemiological, and phenotypic outcomes. The combination of these approaches
with viral outcomes is important in efforts to develop an effective and safe ASFV vaccine.

The study found that African swine fever virus (ASFV) has undergone changes in several MGF gene
families, which may contribute to its success in replicating. Domain-based classification was used,
eliminating the need for genome alignment, which can be difficult due to variations in MGFs among
different ASFV strains. Further investigation is recommended.

The study suggests that using profile HMM domain scores for hierarchical clustering can identify
differences in encoded proteins between similar genomes, complementing standard phylogenetic
approaches. The method was applied to three sets of ASFV genomes and identified differences in MGF
genes as well as potential virulence factors. The computational tools for this analysis are available as a
Docker image with instructions. The study was funded by several organizations, and there were no
conflicts of interest.