[dpryan]

Why would bwa try to pair two reads that have different names just because they are on the nth read in a list? I'm having the same problem, with my pe read names not sorting to same line. I traced this problem to my quality trimming, but I don't understand why you would throw away a perfectly good read just because its mate didn't pass quality trimming. This is the only way you could have nth reads lining up after trimming.

Because reads in a pair needn't have the same name. In addition, it would absolutely kill performance to have the reads just randomly sorted in the files such that bwa would need to constantly search for the position of each mate. All aligners that I've ever seen assume that Nth read in each file are paired for that reason.

You needn't throw the unpaired reads away. Just write them to a separate fastq file (many trimmers will have an option to do this for you).

[etwatson]

That makes sense from a performance point of view. I began to write a script to save the unpaired(orphaned) reads, and it looks like there are plenty of scripts out there that ensure proper pairing of reads.

Thanks.