I am using your lab database AlleleDB for allelic imbalance SNPs of ChIP-seq data and I really find it a formidable resource for further research.
I noticed that many of the significant SNPs for allelic imbalance are SNPs that have only the reference allele in the ChIP-seq data. See the first 10 lines as an example from your file ASB.auto.v2.1.aug16.txt.tgz that can be downloaded from AlleleDB database :
chr start end TF_indiv_ASB ref alt cA cC cG cT p.binomial p.betabinomial
chr1 91604 91605 SA1_NA19099_ASB C T 0 45 0 0 5.68E-14 0.000376021311925
chr1 714018 714019 POL2_NA18505_ASB A G 2 0 19 0 0.000221252441406 0.003114747121148
chr1 714018 714019 POL2_NA19099_ASB A G 2 0 19 0 0.000221252441406 0.002559182427128
chr1 762484 762485 RPB2_NA11894_ASB C A 0 11 0 0 0.0009765625 0.004566586027278
chr1 762600 762601 RPB2_NA11894_ASB T C 0 12 0 0 0.00048828125 0.002968280917731
chr1 840035 840036 SA1_NA18951_ASB T G 0 0 1 168 2.44E-37 0.001745366786663
chr1 937397 937398 SA1_NA19099_ASB G A 0 0 81 0 8.27E-25 5.03E-05
chr1 1167795 1167796 RPB2_NA11894_ASB C T 0 26 0 0 2.98E-08 2.29E-05
chr1 1407371 1407372 RPB2_NA11831_ASB T C 0 0 0 41 9.09E-13 0.000260638330353
chr1 1551110 1551111 POL2_NA19099_ASB C G 0 30 0 0 1.86E-09 1.67E-06
If I understand well allelic imbalance, only SNPs with genotypes that are heterozygous are tested for allelic imbalance.
In this case, the first SNP in the above table (for instance) has a genotype C/T, but shows 45 reads with C and 0 reads with T. This is indeed a strong allelic imbalance. In your data, there are many examples like that.
However, I heard many biologists saying that cell lines are not stable and mutations accumulate over time.
Is it possible that the first SNP is actually homozygous C/C in the cells used for ChIP-seq experiment because of mutations that occurred during cell culture, given that the genotyping by 1000 Genome project and the ChIP-seq experiment might have been done at different times or by different labs?
What do you think?
Thank you for your interest in AlleleDB!
Right, only heterozygous SNVs are tested for the imbalance. We used the 1000G Project variants for these analyses, did not call them ourselves. Indeed, allelic calculations are very sensitive to genotype calls. Thus, a false-positive heterozygous SNV call, as well as a mutation in the cell-line as you described (not sure if there can be many of those), could lead to a false allelic imbalance call. If this is crucial for your analyses, a simple strategy would probably be filtering the call-set to keep only the SNVs where the ‘weaker’ allele is supported by at least one read, though this would remove some true strong imbalance calls as well.