RNASEQ preprocessing and insert size

Question

[Omics data mining]

Dear all,

I have RNASeq data. I have to estimate insert size for multiple samples. For this, I am using the CollectInsertSizeMetrics picard application. Here I have to provide input of BAM file (reads mapped to reference). I am confused whether I should do preprocessing of data and then estimate insert size or I should directly move with raw data to estimate insert size. please look into following points/conditions....

1. consider all the reads for mapping and then calculate insert size ?
2. perform quality filtration,adapter removal, mapping and then estimate the insert size ?
3. only perform the adapter removal and then estimate the insert size ?

Every step will make the difference. I just want to follow the accurate way to estimate insert size.

Waiting for reply

Thank you in advance

[Sean Davis]

You might want to take a look at this blog: https://mikelove.wordpress.com/2016/09/26/rna-seq-fragment-sequence-bias/ and at the alpine bioconductor package.

Answer 1

[Devon Ryan]

If you're going to filter the data anyway then filter it before performing the estimation. You care about the insert size of the data you'll actually use. Do track how much data is filtered, though, since that will also indicate whether there was an issue during library prep.

BTW, you might need to align against the transcriptome to get an accurate number.

Answer 2

[GenoMax]

If all you are interested in is finding insert size then use BBMap suite like this. There are two ways you can estimate the insert size.