Abstract
A large portion of the information contained in next-generation sequencing data is potentially lost through classical bioinformatics analysis. Both the mapping of sequencing reads to a genome or transcriptome and filtering results to focus on known gene regions eliminate useful information. This is especially true in cancer studies where patient transcriptomes or genomes may vary from their references.We created a novel approach that makes use of recent advances in genetic algorithms, neural networks and feature selection to comprehensively explore massive volumes of sequencing data to classify samples without these biases. Our approach, called GECKO for GEnetic Classification using k-mer Optimisation maximizes the sequencing information used when trying to explain the difference between 2 or more samples. Our algorithm has been effective at classifying data from large-scale cancer studies using mRNA-seq, circulating DNA or whole-genome resequencing.iMOKA (interactive multi-objective k-mer analysis) is a software that enables the comprehensive analysis of sequencing data from large cohorts to generate robust classification models or explore specific genetic elements associated with disease etiology. iMOKA uses a fast and accurate feature reduction step that combines a Naïve Bayes classifier augmented by an adaptive entropy filter and a graph-based filter to rapidly reduce the search space. By using a flexible file format and distributed indexing, iMOKA can easily integrate data from multiple experiments and also reduces disk space requirements and identifies changes in transcript levels and single nucleotide variants.Our software could be run on a desktop computer and enable scientists and clinicians to discover novel informative sequences in their own NGS data.Accurate quantification and detection of intron retention levels require specialized software. Building on our previous software, we have created a suite of tools: IRFinder-S, to analyse and explore intron retention events in multiple samples. Specifically, IRFinder-S allows a better identification of true intron retention events using a convolutional neural network, allows the sharing of intron retention results between labs, integrates a dynamic database to explore and contrast available samples and provides a tested method to detect differential levels of intron retention.