Can You Do Some Recommendations For An Indel/Snp/Cnv Simulator?

Question

[Pascal]

Hi

I'm writing a piece of software to simulate Structural Variations in Genome. So far, I have written a first version with a simple set of features:

• accept a FASTA in input,
• write the new genome into a new FASTA file and the variations into a VCF file,
• simulate SNP,
• simulate indels,
• the frequency of SNP, Insertion and Deletions are configurable.

Before going ahead with new features, I would like to hear from your feedback and recommendations.

I thought that one of the possible improvements (to make it more "real") could be to be able to control the frequency of variants depending on the region of the chromosome: to generate for instance more variants in non-coding regions.

Thanks in adance for any input!

Answer 1

[Pierre Lindenbaum]

some trivial suggestions:

• more SNPs in the introns, non-coding regions
• more SNPs in the less conserved regions ( see the GERP score)
• more synonymous snps in the exons.

[Pascal]

Thanks for your answer. To know if my REF base is whether in intron or exon I plan to use BED files generated from the USCS Table tool. Do you think it is a good idea?

[Pascal]

BTW Pierre, in order to generate synonymous SNPs I have to read groups of 3 nucleotides from genome (then see if a base change will be synonymous or not) AND to detect also frame-shift. Is there anything else I should take care of if I decide to implement simulation of sSNP?

[Pierre Lindenbaum]

you could use the human codon frequencies ?

Answer 2

[lh3]

To the best of my knowledge, no one knows how to simulate CNV/SV/indel to an acceptable accuracy. If you really want to get a good simulation, you can simulate reads from a de novo assembly and then get the positions of simulated reads from an assembly-to-assembly alignment. This procedure is quite complicated, but is closest to the truth.

[Pascal]

Thanks for your answer. It is interesting but for the time being I just can't invest too much time :-S I'm looking for a trade-off between complexity of the algo and getting closer to the reality.