Aligning Mb length contigs to a reference genome

Question

[sanjitsbatra]

I am using Bowtie2 for aligning contigs with N50 ~5Mb to a reference. I am getting the following error:

bowtie2-align died with signal 6 (ABRT) (core dumped)

I am using a machine with 100 GB RAM, and 24 cores, so I think memory should not be an issue. Is there any fix for this issue?

Alternatively, I am considering BWA-mem, BLAST, BLAT, SOAP2 and Mummer for the above task. Which one of these is most appropriate, given that I prefer an end-to-end alignment for each contig, instead of an accurate local chimeric alignment?

Thanks!

[Rohit]

I probably saw this too late, probably helps someone looking for an alternative now.
bwasw/bwamem were designed for this particularly, so it is pretty fast.. Speed of blat would decrease when the query size is more than 200kb. Blast would be highly accurate, but slow. Mummer should be able to do it (I have no experience with it). Lastz would be accurate too but speed would be low, and it is not for multiple cores.

[User000]

this is helpful for me, however I was wondering if I use blast, if gives me alignments as blocks, what if I need to obtain a general similarity % coverage and score? also the same for bwa mem, is does not give me this kind of statistics if I am not mistaken? Thanks in advance

[Rohit]

The similarity %, coverage and score must be easy to do based on alignment hits. You could look for calculate these by yourself or use extended version of blast with extended columns. Something like

blastn \
  -task megablast \
  -query query.fa \
  -db database \
  -out hits \
  -perc_identity 90 \
  -outfmt "6 qseqid sseq id pident length mismatch evalue bitscore qcovs"

Answer 1

[thiagomafra]

Try using the software CONTIGuator. Take a look in More Info (in green). I used the web version as well as Command-line version with success.

[sanjitsbatra]

I tried the contiguator. However, since it uses blast+'s blastn within it, it turns out to be very slow(>24 hours) for contigs of one human chromosome. Is there a faster and/or better alternative?