A bioinformatic filter for improved base-call accuracy and polymorphism detection using the Affymetrix GeneChip® whole-genome resequencing platform

Pandya, Gagan A.; Holmes, Michael H.; Sunkara, Sirisha; Sparks, Andrew; Bai, Yun; Verratti, Kathleen; Saeed, Kelly; Venepally, Pratap; Jarrahi, Behnam; Fleischmann, Robert D.; Peterson, Scott N.

The effects of each filtering step on the base-calling accuracy and thus the number of true and false-positive SNPs in the filtered set are shown in Table 3 . The low homology filter eliminated between 44 and 89% of the false positive SNP calls that were in the unfiltered set, while eliminating fewer than 1% of the true positives in all the experiments. The footprint effect filter had a much larger impact on the SCHU S4 results than on the LVS results. This is not surprising since this filter eliminates spurious SNP calls that occur because of genuine SNPs in the experimental DNA sample, and the SCHU S4 sample contains a substantially larger number of genuine SNPs compared with LVS. Each successive filter further reduced the set of false-positive SNP calls with a cost in terms of a loss of some true-positive SNPs. The cumulative effect was a reduction of the initial false-positive set by over 98% in the case of LVS and over 91% in the case of SCHU S4, with no loss of true positives for LVS and a loss of about 10.7% of true positives for SCHU S4 after implementation of the filters. The overall base-calling accuracy (considering both reference calls and SNP calls) was 99.999% for LVS and 99.992% for SCHU S4. These accuracy rates are equivalent to Phred quality scores of 50 and 41, respectively.