Read alignment for ChIPseq

Question

[AlexAbdulkaderKheirallah]

Dear all,

Is there some consensus in terms of preferred parameters for read alignment for ChIPseq analysis? I'm particularly interested on whether it is preferable to discard reads to map to more than one position in the genome. This is what I did in my analysis but I was wondering whether it is better to keep even the reads that map >10 locations in the genome? Is PWM motif probability improved if more reads are retained? I would have thought so since PWM is inferred from reads dataset?

Thanks

[GouthamAtla]

Its not good to keep reads that map to more than 10 locations. The PWM is based on number of observations at each position. Keeping multi mapped reads may introduce bias in calculations.

Answer 1

[Manvendra Singh]

Its always better to keep one alignment per read.

for this you could go for --best --strata functions from bowtie

if the read is mapping to more than 10 locations then probably its coming from repetitive elements. I would not suggest to discard them because lot of repetitive elements provide transcription factor binding sites for neighbor genes. but On the contrary, there would also be chance of false positivity.

better to take one alignment per read, allow multihits.

after mapping you could filter them out for downstream analysis if they are not informative for your questions