recount-brain: a curated repository of human brain RNA-seq datasets metadata


Leonardo Collado-Torres 6,*


Ashkaun Razmara1 Shannon E Ellis2 Dustin J Sokolowski3 Sean Davis4 Michael D Wilson3 Jeffrey Leek5 Andrew E Jaffe5,6

1 Frank H. Netter MD School of Medicine at Quinnipiac University, North Haven, CT
2 Department of Cognitive Science Department, University of California San Diego, La Jolla, CA
3 Department of Molecular Genetics, University of Toronto
4 Center for Cancer Research, National Cancer Institute, NIH
5 Department of Biostatistics, Johns Hopkins Bloomberg School of Public Health, Baltimore
6 Lieber Institute for Brain Development, Johns Hopkins Medical Campus, Baltimore

1 Introduction

  1. Uniformly-processed RNA-seq is available in recount2 (Collado-Torres et al. 2017) and other projects;
  2. Sample metadata from SRA is inconsistent, thus re-using this public data is challenging;
  3. Metadata can be predicted from expression (Ellis et al. 2018) and mapped to ontologies (Bernstein, Doan, and Dewey 2017).

2 Methods

We identified SRA studies present in recount2 that had at least 4 samples with at least 70% of them were predicted to correspond to the brain using phenopredict (v0.0.03) (Ellis et al. 2018). Figure 6 of (Razmara et al. 2019) shows the reproducible curation workflow we followed that briefly involved: creating a list of metadata variables of interest, documenting which part of the paper/supplement the information came from, and any custom modifications. We merged recount-brain with GTEx and TCGA brain sample metadata and linked to controlled vocabulary terms for Brodmann region, tissue and disease.

3 Results

In total, there are 6,547 samples with metadata in recount-brain with 5,330 (81.4%) present in recount2 from 62 SRA studies, GTEx (n=1,409) and TCGA (n=707). The curated metadata can be interactively explored through jhubiostatistics.shinyapps.io/recount-brain/. Figure 3.1 exemplifies some of the metadata information available for these studies.

Overview of some recount-brain sample metadata variables

Figure 3.1: Overview of some recount-brain sample metadata variables

3.1 Example usage

Select studies or add the sample metadata to the expression data with recount::add_metadata() (Figure 3.2).

Access recount-brain using the recount Bioconductor package

Figure 3.2: Access recount-brain using the recount Bioconductor package

As an example of how you can use recount-brain, we used studies with post mortem interval (PMI) information to assess whether expression of RNASE2 is associated with PMI. In studies present in recount-brain we did find an overall association as shown in Figure 3.3 in contrast to (Ferreira et al. 2018)’s findings. A sensitivity analysis releaved study variability which is why Ferreira et al likely did not observe this association.

Replicate findings from other studies using recount-brain

Figure 3.3: Replicate findings from other studies using recount-brain

We used recount-brain to determine the consistency of gene variability across glioblastoma studies SRP027383 and SRP044668 as well as TCGA (Figure 3.4).

Assess consistency of gene variability across glioblastoma studies

Figure 3.4: Assess consistency of gene variability across glioblastoma studies

4 Conclusions

  1. recount-brain (Razmara et al. 2019) facilitates human brain RNA-seq analyses.
  2. recount-brain can be used for reproducing analyses, replicating findings and assessing cross-study variability.
  3. Curation efforts are complementary to prediction efforts (Ellis et al. 2018) and automatic ontology mapping (Bernstein, Doan, and Dewey 2017).
  4. Our reproducible curation workflow can be adapted to curate more samples and other studies.

References

Bernstein, Matthew N., AnHai Doan, and Colin N. Dewey. 2017. “MetaSRA: Normalized Human Sample-Specific Metadata for the Sequence Read Archive.” Bioinformatics 33 (18): 2914–23. https://doi.org/10.1093/bioinformatics/btx334.

Collado-Torres, Leonardo, Abhinav Nellore, Kai Kammers, Shannon E. Ellis, Margaret A. Taub, Kasper D. Hansen, Andrew E. Jaffe, Ben Langmead, and Jeffrey T. Leek. 2017. “Reproducible RNA-Seq Analysis Using Recount2.” Nature Biotechnology 35 (4): 319–21. https://doi.org/10.1038/nbt.3838.

Ellis, Shannon E., Leonardo Collado-Torres, Andrew Jaffe, and Jeffrey T. Leek. 2018. “Improving the Value of Public RNA-Seq Expression Data by Phenotype Prediction.” Nucleic Acids Research 46 (9): e54–e54. https://doi.org/10.1093/nar/gky102.

Ferreira, Pedro G., Manuel Muñoz-Aguirre, Ferran Reverter, Caio P. Sá Godinho, Abel Sousa, Alicia Amadoz, Reza Sodaei, et al. 2018. “The Effects of Death and Post-Mortem Cold Ischemia on Human Tissue Transcriptomes.” Nature Communications 9 (1): 490. https://doi.org/10.1038/s41467-017-02772-x.

Razmara, Ashkaun, Shannon E. Ellis, Dustin J. Sokolowski, Sean Davis, Michael D. Wilson, Jeffrey T. Leek, Andrew E. Jaffe, and Leonardo Collado-Torres. 2019. “Recount-Brain: A Curated Repository of Human Brain RNA-Seq Datasets Metadata.” bioRxiv, April, 618025. https://doi.org/10.1101/618025.

Analysis-ready human curated sample metadata for brain RNA-seq studies