Figure 1

Sources of error in read alignments and allele-specific read counts contribute to bias in estimation of ASE and AI. Here we consider error originating from sequence similarity in the genome (e.g., repeats and duplications) and hidden variation (missed or false SNPs). The examples shown illustrate cases for alignments to a single reference (A-C) and to multiple references (D-E). Alignments to augmented references are expected to behave similarly to alignments to multiple references. A) Masking SNPs located in regions with strong sequence similarity to other locations in the genome (genome sequence ambiguity) can result in alignment error, the best match in the masked reference may be located in a location other than the true source of the read. B) Algorithms that account for multiple mapping can result in allele bias when reads from one of the alleles are discarded or are mapped randomly, while reads from the other allele map to their true source location. C) For a single unmasked reference, reads from one of the alleles may not align at all, resulting in bias toward the other allele. D) When two references are used (one for each parental genome), differences between the references in genome sequence ambiguity can result in allele bias for the same reason as outlined in B. E) Sequencing errors in one reference can result in allele bias when reads from both (identical) alleles align best to the other reference.