The Soybean Expression Atlas v2 A Comprehensive Database of Over 5000 RNA-seq Samples

bioRxiv preprint doi: https://doi.org/10.1101/2023.04.28.538661; this version posted April 29, 2023.
The copyright holder for this preprint

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is
made available under aCC-BY 4.0 International license.
The Soybean Expression Atlas v2: a comprehensive database of over

5000 RNA-seq samples
Fabricio Almeida-Silva1,2*, Francisnei Pedrosa-Silva3, Thiago M. Venancio3*
1
Department of Plant Biotechnology and Bioinformatics, Ghent University, 9052 Ghent, Belgium.
2
VIB Center for Plant Systems Biology, VIB, 9052 Ghent, Belgium.
3
Laboratório de Química e Função de Proteínas e Peptídeos, Centro de Biociências e Biotecnologia,
Universidade Estadual do Norte Fluminense Darcy Ribeiro, Campos dos Goytacazes, Brazil.
*
To whom correspondence should be addressed (fabricio.almeidasilva@psb.vib-ugent.be;
thiago.venancio@gmail.com).
bioRxiv preprint doi: https://doi.org/10.1101/2023.04.28.538661; this version posted April 29, 2023. The copyright holder for this preprint
ABSTRACT
Soybean is a crucial crop worldwide, used as a source of food, feed, and industrial products
due to its high protein and oil content. Previously, the rapid accumulation of soybean
RNA-seq data in public databases and the computational challenges of processing raw
RNA-seq data motivated us to develop the Soybean Expression Atlas, a gene expression
database of over a thousand RNA-seq samples. Over the past few years, our database has
allowed researchers to explore the expression profiles of important gene families, discover
genes associated with agronomic traits, and understand the transcriptional dynamic of
cellular processes. Here, we present the Soybean Expression Atlas v2, an updated version
of our database with a 4-fold increase in the number of samples, featuring transcript- and
gene-level transcript abundance matrices for 5481 publicly available RNA-seq samples. New
features in our database include the availability of transcript-level abundance estimates and
equivalence classes to explore differential transcript usage, abundance estimates in
bias-corrected counts to increase the accuracy of differential gene expression analyses, a
new web interface with improved data visualization and user experience, and a reproducible
and scalable pipeline available as an R package. The Soybean Expression Atlas v2 is
available at https://soyatlas.venanciogroup.uenf.br/, and it will accelerate soybean research,
empowering researchers with high-quality and easily accessible gene expression data.
Keywords: transcriptomics, bioinformatics, gene expression, integrative biology.

Introduction
Soybean (Glycine max [L.] Merr.) is one of the most important crops worldwide due to its
high protein and oil content, and it is used as a source of food, feed, and industrial products.
The availability of its genome sequence (Schmutz et al., 2010) and advances in RNA
sequencing have revolutionized our understanding of the molecular mechanisms underlying
important traits, such as yield (Lu et al., 2021; Yang et al., 2021), stress tolerance (Tamang
et al., 2021; Wang et al., 2021), and seed composition (Jones and Vodkin, 2013; Lu et al.,
2016). RNA-seq data of soybean are accumulating at an unprecedented rate in public
databases, enabling researchers to access a vast amount of transcriptomic information for
gene expression analysis and functional studies (Almeida-Silva et al., 2021). However,
obtaining gene-level transcript abundance (i.e., “gene expression”) matrices from raw
RNA-seq data is time consuming and requires heavy computational resources.
To address this issue, we previously developed the Soybean Expression Atlas, a
gene expression database comprising 1298 publicly available RNA-seq samples that were
systematically collected, processed, and quantified using a unified pipeline (Machado et al.,
2020). Since its release, the Soybean Expression Atlas has allowed researchers to explore
the expression profiles of important gene families (Almeida-Silva and Venancio, 2022;
Coelho et al., 2022; Hou et al., 2022; Huang et al., 2021; Sangi et al., 2021), discover genes
associated with agronomic traits (Zhang et al., 2022; Almeida-Silva and Venancio, 2021;
Almeida-Silva and Venancio, 2023a; Turquetti-Moraes et al., 2022; Guo et al., 2022), and
understand the dynamics of gene regulation underlying plant physiology (Almeida-Silva et
al., 2020; Sangi et al., 2023; Chen et al., 2023; Li et al., 2021). Up to April 2023, our
database has been accessed by more than ten thousand users, 60% of which are recurring.
However, a vast amount of new RNA-seq samples have been deposited in public databases
since then, creating an urge for an updated version of our database.
Here, we present the Soybean Expression v2
(https://soyatlas.venanciogroup.uenf.br/), an updated version of our database comprising
transcript- and gene-level transcript abundance matrices for 5481 publicly available
RNA-seq samples. Besides the 4-fold increase in number of samples, the new features of
the Soybean Expression Atlas v2 include: i. the availability of transcript-level abundance
estimates and equivalence classes, which can be used to explore differential transcript
usage; ii. abundance estimates in bias-corrected counts, which corrects for technical biases
in raw read counts, increasing the accuracy of differential gene expression analyses; iii. a
new web interface with significant improvements in data visualization and user experience,
and; iv. a reproducible and scalable pipeline available as an R package. The Soybean
Expression Atlas v2 will keep accelerating soybean research, empowering researchers with
high-quality and easily accessible gene expression data.
Materials and Methods

Reference genome data acquisition
All genomic data used in this version of the Soybean Expression Atlas were based on the
Wm82.a4.v1 assembly of the Glycine max genome. The reference transcriptome and
transcript-to-gene ID mappings were downloaded from PLAZA Dicots 5.0 (Van Bel et al.,
2022).
Obtaining transcript abundances from raw sequencing data

To make our pipeline reproducible, we developed an R package named bears
(https://github.com/almeidasilvaf/bears) that integrates multiple RNA-seq software tools,
allowing users to estimate transcript- and gene-level transcript abundances from raw RNA
sequencing data available on NCBI’s Sequence Read Archive (SRA) (Sayers et al., 2023).
The package supports transcript abundance estimation using both alignment-based
approaches (i.e., consisting of aligning reads to a reference genome followed by
quantification) and lightweight mapping approaches (i.e., direct abundance estimation
through pseudomapping or selective alignment to a reference transcriptome). In this version
of the Soybean Expression Atlas, we only used the lightweight mapping approach, as it is
dramatically faster than alignment-based approaches while also accounting for technical
biases affecting transcript quantification (’t Hoen et al., 2013; Patro et al., 2017).
Metadata for all RNA-seq experiments available on the SRA (up to June 2022) were
obtained with the function create_sample_info() and filtered to remove experiments using
SOLiD sequencing (due to its obsolescence) and long-read technologies (which are typically
used for transcript assembly, not for quantification). FASTQ files were downloaded from the
European Nucleotide Archive’s mirror of the SRA using the function download_from_ena(),
and file integrity was checked with the check_md5() function. Adapters and low-quality
bases were identified and removed with fastp (Chen et al., 2018) using the function
trim_reads(). FASTQ files with mean read length <40 and/or Q20 rate <80% after filtering
were considered as having insufficient quality and removed. Transcript abundances for the
samples that passed the sequence quality check were estimated with salmon (Patro et al.,
2017) using the function salmon_quantify(), which performs mapping to a reference
transcriptome followed by quantification. Finally, we removed samples for which <50% of the
reads failed to map. Salmon was run with the ‘--dumpEq’ flag to generate equivalence
classes that can be used in differential transcript usage analyses.
Summarizing transcript-level to gene-level abundances

Gene-level transcript abundances were obtained with the function salmon2se(), which reads
quant.sf files created by salmon and creates a SummarizedExperiment object containing
gene-level abundances in transcripts per million (TPM) and read counts. The count matrix is
corrected for biases using the “bias correction without an offset” method implemented in the
tximport package (Soneson et al., 2015), which performs scaling using the average
transcript length over samples and library size.
Dimensionality reduction
To maximize biological signal and reduce noise, we only used genes with highly variable
transcriptional patterns for dimensionality reduction. We used the scran package (Lun et al.,
2016) to model the mean-variance relationship in a matrix of log-transformed counts
normalized by library size, followed by the extraction of the top 5000 genes with the highest
biological components (Supplementary Figure S1A). We performed a principal component
analysis (PCA) of the highly variable genes and extracted the top 8 principal components to
perform dimensionality reduction with t-stochastic neighbor embedding (t-SNE) and uniform
manifold approximation and projection (UMAP) (Supplementary Figure S1B). As t-SNE and
UMAP results can vary depending on perplexity values and number of neighbors,
respectively, we ran both algorithms with 6 different values (10, 20, 30, 40, 50, and 60) and
selected the optimal value based on visual inspection. The optimal perplexity value for the
t-SNE was 60, and the optimal number of neighbors for the UMAP was 30.
Construction of the web application

The web application is a Shiny app that uses the shinydashboard template (Chang et al.,
2021; Chang and Ribeiro, 2019). The gene expression database used in the app is stored in
a partitioned parquet directory that can be downloaded from the FigShare repository
associated with this manuscript (Almeida-Silva and Venancio, 2023b). The interface between
R and the Apache Arrow platform is performed with the arrow R package (Richardson et al.,
2021). Data visualization is powered by ggplot2 (Wicham, 2016), ComplexHeatmap (Gu,
2022), and plotly (Plotly Technologies Inc, 2015).
Results and Discussion

The Soybean Expression Atlas comprises 5481 high-quality samples
We downloaded 6921 RNA-seq samples available on the European Nucleotide Archive
(ENA) and processed them in a unified pipeline to generate a large database of gene-level
transcript abundances (Figure 1A, see Materials and Methods for details). After trimming
reads to remove adapters and low-quality bases, we performed the first quality check, which
consists in removing samples with mean read length <40 bp and/or Q20 rate <80%. 158
samples did not pass this quality check and were removed (Figure 1B, Supplementary Table
S1). Further, after mapping reads to the reference transcriptome, we performed the second
checkpoint, consisting of removing samples with mapping rates <50%. This step led to the
removal of 1282 samples (Figure 2B, Supplementary Table S2), resulting in the final
Soybean Expression Atlas v2 with 5481 samples (Supplementary Table S3).
To have a more detailed view of the quality metrics, we explored the distributions of
sequence quality and mapping statistics. We observed that only a small fraction of the
samples had mean read lengths <40, and nearly all of the samples had Q20 rates >80%,
indicating that the removal of 158 samples in the first quality checkpoint was mainly due to
reads being too short (Figure 1C and 1D). The distribution of mapping rates revealed that
most of the samples had high mapping rates, but two small peaks can be seen around rates
of 35% and 0% (Figure 1E). These samples with low mapping rates represent experiments
that: i. aim at quantifying abundances of non-coding regions (e.g., small RNAs) or; ii.
inoculate soybean tissues with microorganisms and aim at quantifying bacterial transcript
abundances within soybean tissues. The median and mean number of reads per sample
were 30 and 39 million reads (Figure 1F). The summary statistics indicate that soybean
RNA-seq data in public databases overall have good quality.
Newly deposited BioSamples reveal the state of the soybean genomics field
We calculated the frequency of samples for each body part and observed that leaf, root,
shoot, seed, and embryo had the largest number of samples, with 1804, 1237, 1020, 500,
and 166 samples, respectively (Figure 2A). Compared to the previous atlas version
(Machado et al., 2020), there was a 12-fold increase in the number of root samples, which
moved from the fourth to the second position in the ranking by number of samples. When
considering only newly deposited samples, the number of root samples was close to the
number of leaf samples (n = 1137 and 1203, respectively), suggesting that root might
overtake leaf in the near future (Figure 2A). Importantly, body parts that had a negligible
sample size in the previous version of the atlas, such as nodule and flower (n = 2 and 18,
respectively), are currently represented by significantly larger sample sizes (n = 63 and 51,
respectively), which allows more meaningful and reliable conclusions from such data.
A considerable amount of newly deposited root samples represent microbial
inoculation experiments with both pathogenic and plant growth-promoting microorganisms,
revealing increasing global efforts by the soybean genomics community to understand
soybean-microorganism associations (Almeida-Silva et al., 2021). Likewise, newly generated
samples comprising seed parts (i.e., seed, embryo, and cotyledon) mostly aim at
understanding oil biosynthesis at the transcriptional level, suggesting that this field of
research has driven much attention in recent years. Collectively, our findings demonstrate
that newly deposited RNA-seq samples can be used as proxies to summarize the current
state of the soybean genomics field.
Transcript abundance differences between body parts explain sample clustering

We performed dimensionality reduction with t-SNE and UMAP using the top 8 principal
components to explore global patterns in transcript abundance similarities (Figure 2B and
2C). We found that samples clustered well by body part in both the t-SNE and UMAP
representations, but the UMAP representation led to denser clusters with larger distances
between them (Figure 2B and 2C). The separation in clusters was unclear for some body
parts, which is due to their close anatomical proximity and similarity. For instance, some
embryo samples were nearly indistinguishable from endosperm samples, and some nodule
samples were very close to root samples (Figure 2B). Likewise, seed samples were highly
heterogeneous, with some samples closer to embryo, and other samples closer to seed coat
(Figure 2B). As seeds are complex organs that comprise multiple tissues, their larger spread
in t-SNE and UMAP representations is unsurprising.
The UMAP and t-SNE representations are consistent with the expected clustering
patterns, with samples grouping by body part. Although some authors argue that
dimensionality reduction with UMAP better preserves the global structure of the data as
compared to t-SNE (Becht et al., 2019), others have demonstrated that these methods
perform equally well, with differences in global structure attributed to initialization choices
(Kobak and Linderman, 2021). However, UMAP has been shown to produce denser and
more compact clusters as compared to t-SNE, with more space between clusters (Kobak
and Linderman, 2021), which is indeed what we observed (Figure 2B and 2C). Despite the
differences between methods, both the t-SNE and UMAP representations reveal that
samples cluster by body part.
Geographic patterns in data deposits and the evolution of sequencing technologies

We calculated the number of samples per country to explore geographical patterns in
soybean RNA-seq data generation. We observed that world leaders in soybean production
(i.e., US, China, Brazil, Canada, and India) are also world leaders in soybean RNA-seq data
generation (Figure 3A and 3B) (http://soystats.com). Strikingly, the number of samples
deposited by institutions in the US and China (n = 3023 and 1405) is much larger than the
number of samples contributed by other countries. For instance, the number of samples
contributed by China is greater than the number of samples contributed by all other countries
combined (excluding the US), and the number of samples contributed by the US is twice as
large as that of China.
Further, we explored summary sequencing statistics to better understand how such
data were generated. Most of the samples were generated with paired-end sequencing (n =
2878), and Illumina HiSeq 2000, HiSeq 2500, and HiSeq 4000 were the most commonly
used sequencing instruments (n = 1674, 1450, and 1300, respectively) (Figure 3D). Sample
deposits over time show that new samples accumulate exponentially (Figure 3C), which
highlights the challenges of maintaining a database such as the Soybean Expression Atlas
and the need for stable funding. The frequency rankings by body part and sequencing
instrument have mostly remained the same over time, but the time-series visualizations
reveal when changes occurred, such as root samples overtaking seed samples in 2018
(Supplementary Figure S2A), the sharp increase in number of samples generated with
Illumina NovaSeq 6000 since 2021 (Supplementary Figure S2C), and paired-end
sequencing overtaking single-end sequencing in 2019 (Supplementary Figure S2B).
A new web interface to keep up with the increasingly complex data sets in databases
We developed a new web interface, available at https://soyatlas.venanciogroup.uenf.br/, to
allow easier and more meaningful search, exploration, and download of the expression data
in the Soybean Expression Atlas v2. To preserve the structure of the previous version, the
new interface comprises 4 pages: Global Expression Viewer, Search gene list, Download by
body part, and Download by project. All pages except Download by body part have been
updated to include features that enhance user experience and interpretation of results. The
web interface allows exploration and download of gene-level transcript abundances, but
transcript-level abundances are also available as SummarizedExperiment objects in the
FigShare repository associated with this publication (Almeida-Silva and Venancio, 2023b).
The Global Expression Viewer allows users to enter a gene ID and visualize mean
and median expression profiles per body part. However, as the Soybean Expression Atlas v2
is highly heterogeneous, with several different developmental stages and treatments for the
same tissue, we now allow users to visualize the expression profile of a gene in each
individual sample by using t-SNE coordinates (Figure 4A). This new feature can help users
detect if the expression of the input gene in a plant part is consistent across all samples or if
there are condition specific-variations (e.g., a gene is highly expressed in water
stress-induced roots, but not in control conditions).
The Search gene list page allows users to enter a gene list (up to 200 genes) and
visualize their expression profiles (in TPM or bias-corrected counts) in selected body parts
as a heatmap. Users can also download the gene expression matrix used to create the
heatmap and sample metadata as tab-separated files. In the latest interface, we updated this
page to include body part annotations and hierarchical clustering of genes in the heatmap,
which facilitates interpretation.
Finally, the Download by project page allows users to download the gene expression
matrix (in TPM or bias-corrected counts) for a particular BioProject. Previously, users could
search projects by DOI or PubMed ID, which restricted download to only projects that have
associated publications. In the latest interface, we included an interactive table of BioProject
metadata with column filters and a search bar that facilitates the search for projects of
interest (Figure 4B). For instance, users can find and download projects containing more
than N samples, projects that contain a body part of interest, or projects that contain a
user-defined text string in their titles or abstracts (e.g., “water stress”, “seed oil”, “disease”,
etc).
Conclusions
In this work, we present an updated version of the Soybean Expression Atlas, a gene
expression database of 5481 high-quality RNA-seq samples publicly available on the NCBI’s
SRA processed with a standardized pipeline. Gene-level transcript abundances in TPM and
bias-corrected counts can be easily explored and downloaded from the web application,
which will keep empowering soybean researchers towards discoveries. Transcript-level
abundances and equivalence classes are also available to allow differential transcript usage
analyses. The easily accessible expression data in our database can be used for various
purposes, such as differential expression analyses, functional studies, and identification of
genes involved in important traits.
Acknowledgements
FA-S acknowledges funding from Ghent University (Methusalem funding,
BOF.MET.2021.0005.01). FP-S and TMV acknowledge funding from Fundação Carlos
Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ; grant

E-26/203.014/2018), Coordenação de Aperfeiçoamento de Pessoal de Nível
Superior—Brasil (CAPES; Finance Code 001), and Conselho Nacional de Desenvolvimento
Científico e Tecnológico. The funding agencies had no role in the design of the study and
collection, analysis, and interpretation of data and in writing.
Data availability
All data and code used in this paper are available in a GitHub repository
(https://github.com/almeidasilvaf/SEA_paper) to ensure full reproducibility.
REFERENCES
Almeida-Silva, F., Moharana, K.C., Machado, F.B. and Venancio, T.M. (2020) Exploring
the complexity of soybean (Glycine max) transcriptional regulation using global gene
co-expression networks. Planta, 252, 1–12.
Almeida-Silva, F., Moharana, K.C. and Venancio, T.M. (2021) The state of the art in
soybean transcriptomics resources and gene coexpression networks. Silico Plants, 3,
diab005.
Almeida-Silva, F. and Venancio, T.M. (2023a) Discovering and prioritizing candidate
resistance genes against soybean pests by integrating GWAS and gene
coexpression networks. Gene, 860, 147231.
Almeida-Silva, F. and Venancio, T.M. (2021) Integration of genome-wide association
studies and gene coexpression networks unveils promising soybean resistance
genes against five common fungal pathogens. Sci. Rep., 11. Available at:
https://www.biorxiv.org/content/10.1101/2021.08.31.458388v1%0Ahttps://www.biorxiv
.org/content/10.1101/2021.08.31.458388v1.abstract.
Almeida-Silva, F. and Venancio, T.M. (2022) Pathogenesis-related protein 1 (PR-1) genes
in soybean: Genome-wide identification, structural analysis and expression profiling
under multiple biotic and abiotic stresses. Gene, 809, 146013.
Almeida-Silva, F. and Venancio, T.M. (2023b) Soybean Expression Atlas v2. ,
14136051083 Bytes. Available at:
https://figshare.com/articles/dataset/Soybean_Expression_Atlas_v2/22448341.
Becht, E., McInnes, L., Healy, J., Dutertre, C.-A., Kwok, I.W., Ng, L.G., Ginhoux, F. and
Newell, E.W. (2019) Dimensionality reduction for visualizing single-cell data using
UMAP. Nat. Biotechnol., 37, 38–44.
Chang, W., Cheng, J., Allaire, J., et al. (2021) shiny: Web Application Framework for R.
2021. R package version 1.6. 0. Ref. Source.
Chang, W. and Ribeiro, B.B. (2019) shinydashboard: Create Dashboards with “Shiny”. R
package version 0.7.1.
Chen, H., Liu, C., Li, Y., Wang, X., Pan, X., Wang, F. and Zhang, Q. (2023) Developmental
dynamic transcriptome and systematic analysis reveal the major genes underlying
isoflavone accumulation in soybean. Front. Plant Sci., 14.
Chen, S., Zhou, Y., Chen, Y. and Gu, J. (2018) fastp: an ultra-fast all-in-one FASTQ
preprocessor. Bioinformatics, 34, i884–i890.
Coelho, F.S., Sangi, S., Moraes, J.L., Silva Santos, W. da, Gamosa, E.A., Fernandes,
K.V.S. and Grativol, C. (2022) Methyl-CpG binding proteins (MBD) family evolution
and conservation in plants. Gene, 824, 146404.
Gu, Z. (2022) Complex heatmap visualization. iMeta, 1, e43.
Guo, Z., Cao, H., Zhao, J., et al. (2022) A natural uORF variant confers phosphorus
acquisition diversity in soybean. Nat. Commun., 13, 3796.
Hoen, P.A. ’t, Friedländer, M.R., Almlöf, J., et al. (2013) Reproducibility of high-throughput
mRNA and small RNA sequencing across laboratories. Nat. Biotechnol., 31,
1015–1022.
Hou, Z., Li, Y., Cheng, Y., et al. (2022) Genome-wide analysis of DREB genes identifies a
novel salt tolerance gene in wild soybean (Glycine soja). Front. Plant Sci., 13.
Huang, J., Li, X., Chen, X., Guo, Y., Liang, W. and Wang, H. (2021) Genome-wide
identification of soybean ABC transporters relate to aluminum toxicity. Int. J. Mol.
Sci., 22, 6556.
Jones, S.I. and Vodkin, L.O. (2013) Using RNA-Seq to profile soybean seed development
from fertilization to maturity. PloS One, 8, e59270.
Kobak, D. and Linderman, G.C. (2021) Initialization is critical for preserving global data
structure in both t-SNE and UMAP. Nat. Biotechnol., 39, 156–157.
Li, Y., Hou, Z., Li, W., et al. (2021) The legume-specific transcription factor E1 controls leaf
morphology in soybean. BMC Plant Biol., 21, 1–12.
Lu, L., Wei, W., Li, Q.-T., et al. (2021) A transcriptional regulatory module controls lipid
accumulation in soybean. New Phytol., 231, 661–678.
Lu, X., Li, Q.-T., Xiong, Q., et al. (2016) The transcriptomic signature of developing soybean
seeds reveals the genetic basis of seed trait adaptation during domestication. Plant
J., 86, 530–544.
Lun, A.T., McCarthy, D.J. and Marioni, J.C. (2016) A step-by-step workflow for low-level
analysis of single-cell RNA-seq data with Bioconductor.
Machado, F.B., Moharana, K.C., Almeida-Silva, F., Gazara, R.K., Pedrosa-Silva, F.,
Coelho, F.S., Grativol, C. and Venancio, T.M. (2020) Systematic analysis of 1298
RNA-Seq samples and construction of a comprehensive soybean (Glycine max)
expression atlas. Plant J., 103, 1894–1909.
Patro, R., Duggal, G., Love, M.I., Irizarry, R.A. and Kingsford, C. (2017) Salmon provides
fast and bias-aware quantification of transcript expression. Nat. Methods, 14,
417–419.
Plotly Technologies Inc (2015) Collaborative data science, Montreal, QC.
Richardson, N., Cook, I., Crane, N., Keane, J., François, R. and Ooms, J. (2021) arrow:
Integration to Apache Arrow.
Sangi, S., Araújo, P.M., Coelho, F.S., Gazara, R.K., Almeida-Silva, F., Venancio, T.M. and
Grativol, C. (2021) Genome-wide analysis of the COBRA-Like gene family supports
gene expansion through Whole-Genome Duplication in soybean (Glycine max).
Plants, 10, 167.
Sangi, S., Olimpio, G.V., Coelho, F.S., Alexandrino, C.R., Da Cunha, M. and Grativol, C.
(2023) Flagellin and mannitol modulate callose biosynthesis and deposition in
soybean seedlings. Physiol. Plant., e13877.
Sayers, E.W., Bolton, E.E., Brister, J.R., et al. (2023) Database resources of the National
Center for Biotechnology Information in 2023. Nucleic Acids Res., 51, D29–D38.
Schmutz, J., Cannon, S.B., Schlueter, J., et al. (2010) Genome sequence of the
palaeopolyploid soybean. Nature, 463, 178–183.
Soneson, C., Love, M.I. and Robinson, M.D. (2015) Differential analyses for RNA-seq:
transcript-level estimates improve gene-level inferences. F1000Research, 4.
Tamang, B.G., Li, S., Rajasundaram, D., Lamichhane, S. and Fukao, T. (2021)
Overlapping and stress-specific transcriptomic and hormonal responses to flooding
and drought in soybean. Plant J., 107, 100–117.
Turquetti-Moraes, D.K., Moharana, K.C., Almeida-Silva, F., Pedrosa-Silva, F. and
Venancio, T.M. (2022) Integrating omics approaches to discover and prioritize
candidate genes involved in oil biosynthesis in soybean. Gene, 808, 145976.
Van Bel, M., Silvestri, F., Weitz, E.M., Kreft, L., Botzki, A., Coppens, F. and Vandepoele,
K. (2022) PLAZA 5.0: extending the scope and power of comparative and functional
genomics in plants. Nucleic Acids Res., 50, 1468–1474.
Wang, K., Bu, T., Cheng, Q., et al. (2021) Two homologous LHY pairs negatively control
soybean drought tolerance by repressing the abscisic acid responses. New Phytol.,
229, 2660–2675.
Wicham, H. (2016) ggplot2: Elegant Graphics for Data Analysis.
Yang, Y., Zhu, X., Cui, R., Wang, R., Li, H., Wang, J., Chen, H. and Zhang, D. (2021)
Identification of soybean phosphorous efficiency QTLs and genes using chlorophyll
fluorescence parameters through GWAS and RNA-seq. Planta, 254, 1–13.
Zhang, Y., Song, J., Wang, L., et al. (2022) Identifying quantitative trait loci and candidate
genes conferring resistance to soybean mosaic virus SC7 by quantitative trait
loci-sequencing in soybean. Front. Plant Sci., 13.
Figures
Fig. 1. Pipeline used to create the Soybean Expression Atlas v2 and summary quality statistics. A. Data
acquisition and processing pipeline used in the Soybean Expression Atlas v2 (see Materials and Methods for
details). Checkpoints represent stages where samples were removed. All data generated at the end of the
pipeline (SummarizedExperiment objects for gene- and transcript-level transcript abundances and equivalence
classes) are available in the FigShare repository associated with this publication. B. Number of BioSamples
before and after each filtering step. Numbers next to the bars represent the remaining samples at each step, and
numbers inside bars represent removed samples relative to the previous step. C. Distribution of mean read
lengths per sample. Dashed red line represents the minimum read length threshold from checkpoint 1. D.
Distribution of Q20 rates per sample. Dashed red line represents the minimum Q20 rate threshold from
checkpoint 1. E. Distribution of mapping rates per sample. Dashed red line indicates the minimum mapping rate
threshold from checkpoint 2. F. Distribution of number of reads (in millions) per sample.
Fig. 2. Frequency of samples per body part and dimensionality reduction. A. Number of samples per body
part. Numbers in bold next to bars represent the number of samples, and numbers in parentheses represent the
increase in number of samples relative to the previous version of the Soybean Expression Atlas. B. t-SNE
representation of samples. t-SNE coordinates were calculated from the top 8 principal components, with
perplexity=60. C. UMAP representation of samples. UMAP coordinates were calculated from the top 8 principal
components, with n_neighbors=30. Principal components used in the t-SNE and UMAP representations were
obtained from a matrix of log-transformed count values for the top 5000 genes with the highest biological
components (see Materials and Methods for details).
Fig. 3. Global distribution of sample deposits and summary sequencing statistics. A. Geographic
distribution of the top 20 countries in number of deposited samples. B. Bar plot of the top 20 countries in number
of deposited samples. Blue bars represent countries that are world leaders in soybean production. C. Cumulative
number of samples over time. Samples that failed the quality checkpoints were not included in the visualization.
D. Summary statistics on library layout and sequencing instruments.
Fig. 4. Example features of the new web application. A. Global Expression Viewer page highlighting the
gene-level transcript abundance profiles of a ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO)
subunit. Upon search, users can visualize transcript abundances based on t-SNE coordinates, a map of plant
body parts, and bar plots of mean and median abundances. B. Download by project page. After selecting a given
BioProject in the interactive table (highlighted in blue), users can download gene expression matrices and/or
sample metadata as tab-separated files.

The Soybean Expression Atlas v2 A Comprehensive Database of Over 5000 RNA-seq Samples

Uploaded by

Copyright:

Available Formats

You might also like

The Soybean Expression Atlas v2 A Comprehensive Database of Over 5000 RNA-seq Samples

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Soybean Expression Atlas v2 A Comprehensive Database of Over 5000 RNA-seq Samples

Uploaded by

Copyright:

Available Formats

bioRxiv preprint doi: https://doi.org/10.1101/2023.04.28.538661; this version posted April 29, 2023.

The copyright holder for this preprint

The Soybean Expression Atlas v2: a comprehensive database of over

Fabricio Almeida-Silva1,2, Francisnei Pedrosa-Silva3, Thiago M. Venancio3

Keywords: transcriptomics, bioinformatics, gene expression, integrative biology.

Materials and Methods

Obtaining transcript abundances from raw sequencing data

Summarizing transcript-level to gene-level abundances

Construction of the web application

Results and Discussion

Transcript abundance differences between body parts explain sample clustering

Geographic patterns in data deposits and the evolution of sequencing technologies

Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ; grant

You might also like