jaclyn-taroni / ref-txome Goto Github PK

The ideas herein were formulated during a trip Kurt Wheeler (@kurtwheeler) and I took to the COMBINE Lab (and the discussion that followed), with input from/discussion with Rob Patro (@rob-p) and Hirak Sarkar (@hiraksarkar), specifically.

The Arabidopsis thaliana example is a rough sketch of what we might expect the RNA-seq component of transcriptomic data processing to look like in its most basic form. (Note that this has not undergone any code review process at present.)

Several A. thaliana RNA-seq samples are downloaded from the European Nucleotide Archive by running 0-athaliana_data_dl.sh in the top directory.

The examples include:

• A paired-end sample with a read length of 100bp -- DRR016125
• A single-end example with a read length of 36bp -- SRR074262
• An single-end experiment, read length 50bp -- PRJNA408323

An A. thaliana custom reference can be obtained by running 1-athaliana_custom_ref.shin the top directory.

This runs scripts/get_custom_transcripts_fasta.sh, which is the main script for preparing the custom reference. The GTF file is first filtered to remove any pseudogenes, as we expect these will negatively affect on our ability to quantify protein-coding transcripts (see also #1), and then passes this filtered GTF and the top level DNA FASTA file to rsem-prepare-reference for building the custom reference. scripts/check_for_pseudogene_removal.R is currently also run to determine if the GTF was correctly filtered and the custom reference contains all the expected transcripts.

The A. thaliana genome information is pulled from Ensembl Plants. There are two other examples: R. norvegicus (from Ensembl) & H. sapiens (from GENCODE), rnorvegicus_custom_ref.sh and hsapiens_custom_ref.sh, respectively.

Two txome indices for use with Salmon (specifically quasi-mapping-based mode) are built by running 2-athaliana_build_index.sh in the top directory:

• A "short" index
  • For use with reads 50bp or less
  • -k 23
• A "normal" index
  • For use with all longer reads (the majority of what is publicly available)
  • -k 31

See the Salmon documentation for more information about building an index for a transcriptome.

Quantification with Salmon for our three examples is performed with 3-athaliana_quant.sh

Note: The distribution of read lengths in a fastq file determines which txome index is appropriate for use with quant.

4-athaliana_tximport.R summarizes the output of salmon quant to the gene-level. Specifically, we are interested in length-scaled transcripts per million (TPM). (See #6.)

An A. thaliana example adapted from Xiajun Dong and Laurent Modolo. (Note that this is not particularly fast.)

zcat athaliana/data/SRR6080007/SRR6080007.fastq.gz \
  | awk 'NR%4 == 2 {lengths[length($0)]++} END {for (l in lengths) \
    {print l, lengths[l]}}'