ATD: Quantitative Methods for Estimating Sequencing Errors

Dana-Farber Cancer Institute, Boston MA

Investigators

Abstract

The investigators propose is to establish a general statistical methodology to estimate error rates for any of the existing second-generation sequencing technologies. They have identified an approach that is broadly applicable, fast, and easy to implement. Important strengths are that it only requires intensity data; it is applicable to the data types that are typically shared publicly; and it does not require the availability of a reference genome or genomes ---a key condition in threat detection applications. Sequencing is a technology to read small words made out of DNA or RNA. In may applications across biology we need to identify words that occur in a book where they are not supposed to be (for example mutations in cancer or pathogens in the intestinal flora). Often these 'bad' words are similar to other words which occur elsewhere in the book, differing only by a letter or two. As seqeuncing is not free of errors, to know whether we are seeing a bad word or a poorly read good word we need to know how easy it is to misread a letter. This proposal is to assess exactly this, with the goal of providing better foundations to all scientific research that uses sequencing.