RNA-Seq Best way to handle multimapping
Hi, I am using STAR to align my illumina reads to a reference. For a multimapped read that maps equally well to multiple locations, should I
- keep all alignments for multiple locations
- keep one alignment (one location, primary alignment) and discard all other alignments (secondary alignments)
- remove all alignments (primary and secondary alignments) for all locations
Which is better (1,2,3) for differential expression analysis [FEATURECOUNTS+DESEQ2] and differential isoform analysis [RSEM]? Should I filter first by MAPQ and then by primary alignment? Also how does one filter a BAM file for primary alignments?
Is it better to change some setting with the aligner during mapping or with the DGE tool?
Thanks.
0 answers
No answers yet.
Log in to answer this question.
More posts like this
-
How does featureCounts handle multimapped reads from Bowtie2 -k 100 in default mode?
written by rachana •
-
Is comparing seeds sufficient, or should alignments be compared instead?
written by curiousfellow@123 •
-
vg giraffe handling of second vs. supplemental/split alignments, novel SVs, and associated MAPQ
written by Thomas •
-
inconsistent results of rsem quantification
written by gnilihzeux •
-
filtering reads that are uniquely mapped
written by Florian Bernard
-
Difference between chimeric alignments and multiple mapping
written by Vanilla
-
BWA MAPQ qualities 0 for primary alignments and discordant reads
written by djomlaart •
-
SAM / BAM alignments
written by John
-
Differential isoform expression analyses
written by firestar
-
Hi-C and multi-mappers
written by int11ap1
How are you planning on getting the counts for DESeq2?
For RSEM, option 1.
FeatureCounts. I have updated the question.
There is likely no unbiased way to select the "best location" out of all locations to which a read maps. Downstream tools (HTseq-count, featureCounts) will ignore multimapping reads by default.
Use Kallisto for expression value and Sleuth to do differential expression. It's really quick and relatively straight-forward.
The featurecounts counts and kallisto estimated counts are similar for me. If I use both output with deseq2, they give similar number of DE genes. But sleuth doesn't seem to give any DE genes.
Are you running kallisto with bootstrapping? Sleuth uses the bootstrap values for dispersion estimation.
Sleuth will not allow you to turn the DE analysis without the bootstrapping, At least in the last version
An alert is given if required
Yes. I do bootstrapping. 100 reps.
Multimapping is one of issues for count-based differential gene expression analysis. Personally I discard all of the multimapping reads by STAR (for human RNA-seq typically should be less than 10%) when counting. Another choice is switching to transcript quantification like kallisto, Salmon, etc.. They can handel the issue of multimapping with statistical model to assign such reads while inferring maximum likelihood estimates of transcript abundances. If you use kallisto, then sleuth maybe the best choice for DGE. Or if you want to stick to count-based tools like DESeq2, edgeR, limma-voom, you can use tximport to convert transcript abundance to counts.