Interpretable CNV-based tumour classification using fuzzy rule based classifiers

DNA copy number variations (CNVs) are widespread structural variants in the genome that are believed to play an important role in tumour evolution. Although CNVs affect a greater fraction of the genome than single nucleotide polymorphism (SNPs), so far they have received much less attention, and their contribution in cancer genomics is not still fully understood. In this work we propose a novel pipeline to support tumour type classification and rule extraction based on somatic CNV data. The pipeline outputs an interpretable Fuzzy Rule Based Classifier (FRBC), on which inference can be made. The pipeline benchmarking is performed over a set of samples of kidney cancer from TCGA. The results show the potential application of the approach: The method is able to classify between three kidney tumour types, with an accuracy of ∼ 93%, using a compact set of ∼ 50 interpretable rules.