Question about indexing the human genome

[chenjy]

Dear all,

I am going to index the human genome (hg19). Besides the sequence file named chr??.fa.gz, there are unplaced-sequence (chr??_gl??????_random.fa.gz) and alternative-haplotype (chr??_?????_hap?.fa.gz) files. Should I index only the base files or all of them? What effect will it have on the alignment results?

Thanks

[GenoMax]

Unless you have a specific need why not download the per-created index files that you can find at the iGenomes site: http://cufflinks.cbcb.umd.edu/igenomes.html

[chenjy]

Unless you have a specific need why not download the per-created index files that you can find at the iGenomes site: http://cufflinks.cbcb.umd.edu/igenomes.html

I am just wondering how much the difference between the alignment results will be by including/excluding those abnormal sequences when building the index.

[GenoMax]

I am just wondering how much the difference between the alignment results will be by including/excluding those abnormal sequences when building the index.

Calling these sequences "Abnormal" may be a bit extreme. These sequences have just not been confidently assigned a location though they are part of the human chromosome/genome.

UCSC has an explanation here: http://genome.ucsc.edu/FAQ/FAQdownloads#download10

If you are only interested in confidently assigned sequences from the genome build then you can exclude them from your analysis.

[chenjy]

Calling these sequences "Abnormal" may be a bit extreme. These sequences have just not been condifently assigned a location though they are part of the human chromosome/genome.

UCSC has an explanation here: http://genome.ucsc.edu/FAQ/FAQdownloads#download10

If you are only interested in confidently assigned sequences from the genome build then you can exclude them from your analysis.

yes, they are actually not abnormal sequences. thanks!

[idonaldson]

If I were creating a new index for a hg19 I would include all sequences, apart from the alternative haplotypes (*hap* files). I would exclude them because the sequences will be highly redundant, which will make mapping reads to those areas problematic particularly if uniquely mapping reads are required.

In general my understanding is that all available sequence should be included in the reference (apart from the caveat above). This is principaly because if your DNA sample produces reads that are present in the unassembled contigs they may erroniously map to the wrong place if the correct sequence is not present in your reference.

If anyone disagrees with this summary i would be very interested in their opinion.