This package contains code to analyze RNA-seq data and generate figures and tables presented in:
A genetic, genomic, and computational resource for exploring neural circuit function
Davis FP*, Nern A*, Picard S, Reiser MB, Rubin GM, Eddy SR, Henry GL.
bioRxiv 2018
Version: updated Aug 27, 2019 to reflect latest preprint version
Contact [email protected] with any questions.
- src - code, organized by language, to turn RNA-seq read files into figures.
- metadata - text tables describing RNA-seq samples
- data - text data files used by code
The data used by this package comes from several sources. We include nearly all data files expected by the R and LSF shell programs, along with README files describing the contents. The only exceptions are large files (eg, genome sequence, gene annotations, transcript sequences, RNA-seq alignment indices), which we do not provide but describe in README files how to obtain or build.
| Data | Source |
|---|---|
| RNA-seq FASTQ files | GEO accession GSE116969 |
| genome sequence | ENSEMBL release 91; based on BDGP6 (dm6) |
| gene structures | ENSEMBL release 91; based on FlyBase 2017_04 |
| FlyBase gene groups | FlyBase 2018_02 |
We also used data from the following papers:
- Rivera-Alaba et al., Curr Biol 2011.
- Takemura et al., Nature 2013.
- Ozkan et al., Cell 2013.
- Konstantinides et al., Cell 2018.
- Davie et al., Cell 2018.
We used the following software on an LSF linux compute cluster.
| # | Software | Source |
|---|---|---|
| 1 | seqtk (2012 Oct 16) | https://github.com/lh3/seqtk |
| 2 | kallisto 0.43.1 | https://pachterlab.github.io/kallisto/ |
| 3 | STAR 2.5.3c | https://github.com/alexdobin/STAR |
| 4 | bedtools 2.15.0 | http://bedtools.readthedocs.io/ |
| 5 | samtools 0.1.19 | https://github.com/samtools/samtools |
| 6 | UCSC bedGraphToBigWig | http://hgdownload.cse.ucsc.edu/admin/exe |
| 7 | picard 1.9.1 | http://broadinstitute.github.io/picard/ |
| 8 | R v3.3.1 | https://www.r-project.org/ |
| 9 | RStan/Stan | http://mc-stan.org/users/interfaces/rstan |
The analysis includes three parts: (1) RNA-seq read processing, (2) modeling gene expression states, and (3) generating figures and tables.
This step is implemented in an LSF script that trims the reads (seqtk), pseudo-aligns them to the transcriptome (kallisto) and aligns them to the genome (STAR), evaluates quality metrics (picard), and generates bigwig tracks for visualization (samtools, ucsc kent tools)
bsub < process_sample.lsf.shWe implemented this step in an R routine that uses the RStan interface to the Stan statistical inference engine to model gene expression states. This step uses an LSF compute cluster for parallelization.
The routine assume 200 cluster nodes with 4 CPUs each (values specified in the setSpecs() routine). On our cluster, this step takes ~ 3 hours in total.
source("../../src/R/analyzeOpticLobeExpr.R")
dat <- main(returnData = TRUE)
runModel(dat, mode="submit")
dat$pFit <- runModel(dat, mode="merge")
dat <- main(dat,returnData = TRUE)
saveRDS(dat, paste0(dat$specs$outDir,"/full_dataset.rds"))An R routine generates all the figures in final form, except for a few where we manually repositioned labels for clarity
makeFigs(dat)