Recalibration of minor alleles in the human reference sequence

Abstract

The intrinsic problem of minor alleles occupying reference positions in the Human Reference Sequence build 37 may challenge the notion of accurate variant calling and result in variant misinterpretation in the clinical practice. In this research study, a bioinformatics pipeline, RecAl, was developed with the primary aim to detect all reference minor alleles and generate three VCF files during sample analysis. These files include the false-positive variants, the false-negative variants, and a separate corrected sample VCF file with the eliminated false-positive variants and incorporated false-negative variants. When the sample files were processed through RecAl, the percentage of false positives variants detected for an alternate allele frequency threshold of 0.90, 0.95 and 0.99 were 9.7%, 7.5% and 5.4% respectively. For the false negative variants, RecAl identified 0.013%, 0.007% and 0.005% respectively. Each of these variants were annotated using popular pathogenicity prediction tools including CADD (Kircher M et al., 2014), Polyphen-2 (Adzhubei I et al., 2010) and SIFT (Ng, P. and Henikoff, S., 2001). From the results, it was presented that 1.24% of the false-positive variants and 0.87% of the false-negative variants are deleterious with significant impact of sequence variation. Additionally, the list generated through RecAl for reference minor alleles was compared to the study carried out by Fuentes F et al., (2012) which focused on false-positive calls due to reference minor alleles in exome regions. From this evaluation, 90% of the variants matched which signifies that the problem of minor alleles occupying reference positions is still prevalent and the list of reference minor alleles generated by RecAl is reliable. Lastly, a comparative analysis of the reference minor alleles in the Human Reference build 37 was compared to the reference minor alleles in build 38 to assess how many reference minor alleles were corrected which resulted in only 9% being corrected.