Eland - GRCh37/hg19 assembly

[AnotherHTS]

Hi All,

I have recently experienced a problem using the human GRCh37/hg19 assembly (from UCSC) with CASAVA/Eland.

After aligning samples with an expected random coverage to the UCSC GRCh37/hg19 assembly, we noticed large areas with zero read coverage. For example no hits were found between positions 1 and
151000000 on chr1.

The fix to this was to place all of the random and unplaced contigs from the assembly into one file before squashing the genome. After we had reduced the number of reference files our alignment ran as expected, and produced the whole genome coverage we expected to see. I was wondering if anyone else has experienced anything similar?

[mgogol]

I just did a run against hg19, but I haven't looked at the data yet.

According to an illumina guy at a user group meeting, you should leave out the haplotype chromosomes when squashing the human genome, otherwise a number of reads will be thrown out for multi-mapping. Also, the genome's too big for squashing if you include them.

[mgogol]

Just an update: Yes, we did see something similar. Ugh.

Re-aligning with a genome directly from Illumina that doesn't include the random and unplaced stuff. I really wish eland squash genomes would give you an error when it fails.