jackhump / cryptex Goto Github PK

Dear Jack!
My name is Alex Rezvykh, I'm a 4th year PhD student from Moscow Institute of Physics and Technology, working in laboratory of Molecular mechanims of biological adaptation. (Moscow,Russia). One of objectives i currently work on is to detect and quantify differential cryptic exons inclusions in RNA-seq data. I'm very interested in usage fo your CryptEx pipeline (https://github.com/jackhump/CryptEx), and i tried to use it with my RNA-seq data by myself (mouse spinal cord, ~25-50 million reads per library, aligned by hisat2, sorted and indexed by samtools). I modifyed some input lines of you cryptex.sh file, and run it in strict mode, but i faced some difficulties running pipeline:

1. Which is the best way to split reference GFF file into inton and exon file? In your code, you mentioned Devon Ryan script, but i cannot found it in your Github repos. I tried to parse GFF by myself (this one
   http://ftp.ensembl.org/pub/release-106/gff3/mus_musculus/Mus_musculus.GRCm39.106.gff3.gz) , using rtracklayer and gread package (construct_introns function), but i got confusing results:

2. Splice extraction, GFF creator and count steps are running just fine, with no errors and creation all necessary BED, GFF files, resulting in identification of ~200 novel exons.

3. I configured your pipeline to run normally till Step4 (forked_dexseq_pipeline_v2.R script), but dexseq counts file looks like below, and dexseq part fails in DEXSeqDataSetFromHTSeq with error:

line 462820 did not have 3 elements
DEXseq counts looking like:
"":"i1" 0
"":"i10" 0
"":"i100" 1
"":"i1000" 0
"":"i10000" 0
"":"i100000" 377
"":"i100001" 28
"":"i100002" 0
"":"i100003" 0

I've never worked with DEXSeq package before, but simple analysis of source code of function DEXSeqDataSetFromHTSeq tells me, that counts obtained are possibly not in the right format.
Summarizing, It seems to me that the roots of the problems are in correct GFF file, and correct way to split it into exons and introns. I would be very grateful to you if you can help me solve this issue.

4. When master script is running, I've got an alert : FUS_dexseq_frame.tab is not a valid set of conditions, and script SJ_analyzer_FUS.sh is empty. It supposed to be a problem with my support file? (i used tab-separated file with header, 1st columnt - unique sample identifier, 2nd - full path to sorted and indexed BAM file, 3nd - FUS dataset, 4th - Condition (TG,WT etc). Please, can you provide to me the right way to create a support file, containing more than 1 conditions, to perform pairwise comparisons later?

Best wishes,
Alex.

• Replace HTSeq with FeatureCounts
• Allow users to submit a GFF of their choosing (GENCODE BASIC/COMP?)
• Reduce number of unnecessary output files
• Allow for easy customising of CryptEx internal settings
• Look into pre-dexseq QC (minimum splice junction counts etc) to reduce burden on dexseq and reduce false positives