BWA and Velvet : Anamolies in contigs and read mapping to reference

[lh3]

What is the percent identity between the contig and the known reference?

[apratap]

Hi All

I believe I have an interesting problem to look at or may be its just me who is experiencing this for the first time.

I have reads from GA for a bacterial genome. I used BWA to map them to known reference. Only 4% reads map back. Seeing the low % I decided to de novo assemble the reads using Velvet .


1. Total of 2150 contigs
2. 601 align to reference using 52 % reads from the original set.
3. k-mer size I used was 29. (read length : 50 bp PE)

I am not sure what to infer from these very contrasting numbers according to me. With Velvet contigs I am able to use 52% of the reads that aligns well to reference whereas with BWA only 4% reads align.

Looking forward for your thoughts.

Thanks!
-Abhi