Newbler mapping: 50% chimeric reads?

[yvan.wenger]

Hello everybody,

I tried to map 454 reads on a genomic reference. I know the genome is not very good in term of contiguity (̃20'000 contigs). I*put below the mapping statistics. I*am a bit surprised, does anyone know the reason why 53% of my reads are considered "chimeric"?

Second question: I made a mapping assembly of illumina reads using tophat/cufflinks/rsem. I would like to do the same reconstruction with my 454 reads now. I would be very grateful if somebody could advise on an efficient tool to do this. Ideally, it would produce an extractable gtf file or transcripts sequences. Of course I do not like the way tophat treat this: by hashing long reads. This is actually why I*tried newbler.

Thanks a lot!

Yvan


readStatus
{
numMappedReads = 1025555, 91.45%;
numMappedBases = 271319305, 87.71%;
inferredReadError = 1.85%, 4919229;

numberFullyMapped = 283828, 25.31%;
numberPartiallyMapped = 115167, 10.27%;
numberUnmapped = 56418, 5.03%;
numberRepeat = 27511, 2.45%;
numberChimeric = 599049, 53.42%;
numberTooShort = 39466, 3.52%;

[tonybolger]

Hello everybody,

I tried to map 454 reads on a genomic reference. I know the genome is not very good in term of contiguity (̃20'000 contigs). I*put below the mapping statistics. I*am a bit surprised, does anyone know the reason why 53% of my reads are considered "chimeric"?

As a WAG, because they cross a contig boundary? This wouldn't be surprising if your assembly isn't good, and exons of one gene are in different contigs.

Second question: I made a mapping assembly of illumina reads using tophat/cufflinks/rsem. I would like to do the same reconstruction with my 454 reads now. I would be very grateful if somebody could advise on an efficient tool to do this.

bwa-sw can do the first bit.