Consensus clustering applied to multi-omics disease subtyping

Abstract Background Facing the diversity of omics data and difficulty selecting one result over all those produced by several methods, consensus strategies have potential to reconcile multiple inputs produce robust results. Results Here, we introduce ClustOmics, a generic clustering tool that use in context cancer subtyping. ClustOmics relies on non-relational graph database, which allows for simultaneous integration both results from various methods. This new conciliates input clusterings, regardless their origin, number, size or shape. implements an intuitive flexible strategy, based upon idea evidence accumulation . computes co-occurrences pairs samples clusters uses this score as similarity measure reorganize into clusters. Conclusion We applied multi-omics disease subtyping real TCGA ten different types. showed is heterogeneous qualities partitions, smoothing reconciling preliminary predictions high-quality clusters, computational biological point view. The comparison state-of-the-art consensus-based tool, COCA, further corroborated statement. However, main interest not compete with other tools, but rather make profit when no gold-standard metric available assess significance. Availability source code, released under MIT license, obtained are GitHub: https://github.com/galadrielbriere/ClustOmics