edgeR library size : UNIQUE reads or all alignments from SAM file?

[carmeyeii]

Hi!

I have what I think is a very basic question.

When you input the table of counts to edgeR, it can either estimate the library sizes based on the total number of counts of all genes in a library, or you can supply this information.

I take this to mean that if the TopHat alignment (the .sam file) assigned a read to two different genes, the total library size calculated by edgeR will be greater than the total mapped [unique] reads (as some reads may have mapped more than once). Whereas if you supply edgeR the library sizes you could choose to give it the total unique mapped reads OR the total number of "mappings".

The question is, which is the accurate thing to use as a "total mapped reads" value: the number of unique reads that mapped (even if they mapped more than once), or the total number of alignments (i.e., counting each instance of every read in the genome?

Thanks for your help,

Carmen

[Gordon Smyth]

If reads can be counted more than once, then the edgeR library size would preferably be the number of unique reads. However, if the count totals are roughly proportional to the number of unique reads, as they probably will be, then it will make little difference to the edgeR analysis which definition you use. Any difference will probably be even more minor after scale normalization.

Gordon