Sort-Seq Tools: sequence-function relationship modeling for massively parallel assays

Abstract

A variety of massively parallel assays for measuring high-resolution sequence-function relationships have been developed in recent years. However, software for learning quantitative models from these data is lacking. Here we describe Sort-Seq Tools, a software package that allows multiple types of quantitative models to be fit to massively parallel data in multiple different ways. We demonstrate Sort-Seq Tools on both simulated and published data from Sort-Seq studies, massively parallel reporter assays, and deep mutational scanning experiments. We observe that, as an inference method, information maximization generally outperforms both least squares optimization and enrichment ratio calculations.