7.8 years ago by
You could try LASTZ.
I didn't understand exactly what you mean by 'link these hits into larger blocks'. But did you specify the -maxIntronSize parameter? Try setting it to 0. However, if you can align your contigs only with these large gaps, you are either doing assembly of RNA-seq and these gaps are real or you might see mis-assemblies. Or maybe your reference sequence is more distant than you thought?
In the light of what you supplied in information, I would simply use blast. Blat was made primarily for highly similar sequences with large insert, e.g. mapping ESTs to the genome, maybe also useful for aligning 454 reads to the genome, but it is not the tool of choice for sequences that diverted 5 million years ago.
Further, there is nothing that suggests that you or blat 'linked' the local alignments in your example, the only link is that these alignments stem from the same query sequence. To assume that there is an alignment spanning the the whole range is not valid, given the parameters of sequence identity and gaps, exactly because of the sequence differences between them. What this also tells you is that there was one large insert.
You could also experiment with gap costs, different substitution costs a bit and also try to set the -maxIntronSize parameter in blat higher, but possibly that wouldn't work either.