ELAND reporting more than 2 mismatches?

[kch]

In our s_N_sorted.txt file, we are finding cases in which ELAND appears to be reporting more than 2 mismatches, which sounds not at all plausible, as far as I understand. For example, we are getting lines that look like this:

HWUSI-EAS548 7 1 60 803 680 0 1 GTGGCGCGCTCCTGTCGTCCCAGCTGCCCGGGAGGTTGAGGCAGGAGAATCGTTTGAACTTGGGAAGTGGAGGGGG a^_`a``Z`[^\^^^^ZX\`\^`^R^Z\\YX\PZTP^\PX^TS[XKUPRPZZRR\\NURT]RSRKS\BBBBBBBBB c10.fa 165114 F 15A11T7CCC6C7C6CC4G7TT1 23

The match descriptor column appears to be reporting an unusual number of mismatches. Can anyone provide any insight on what could be going on? We are using 76 bp reads.

Thanks!

[simonandrews]

I believe the absolute limit on number of mismatched bases only applies within a limited region at the start of the read (24bp?). After that the matching is more relaxed and allows further mismatches. Only allowing 2 mismatches in a 76bp read would result in the rejection of quite a lot of your sequences.

You can probably adjust the matching parameters used by Eland, or use another mapper which allows you to set more specific match parameters.

[kch]

Thanks for your comments, simonandrews!

[lh3]

At least a year ago, Eland was using the first 32bp as seed. It outputted up to by default 10 equally best hits for each read and discarded reads with more than 10 hits. Pairing and extension are both based on these 10 hits only. You can imagine that when the first 32bp are highly repetitive but the last 32bp are unique, eland will miss the hit.

I do not know how eland works in the latest GApipeline.