Machine Learning revolutionizes modern cancer research by personalizing cancer treatment through identification of blood based biomarkers that allow to quickly determine whether a given tumor is malignant or benign at a low cost. This talk presents a novel biomarker identification concept that is based on a Support Vector Machine (SVM) and termed ‘Alikeness estimation by Regression on Transcriptome DECOnvolution’ (ART DÉCO).
As first of its kind, ART DÉCO identifies biomarkers suitable for malignancy estimation of rare neuroendocrine pancreatic cancer by deconvolution of the genomic data of tumors and carcinomas. The deconvolution is performed by a SVM Regression that allows to quantify the similarity of tumor samples to samples with known and blood detectable biomarker signatures. Initially, the deconvolution concept will be explained and subsequently elaborated how a SVM successfully establishes a similarity metric over genomic landscapes.
Advantages and limitations of the SVM approach will be discussed and benchmarks and congruence of the ART DÉCO predictions with clinical data be shown to eventually conclude with an outlook on an alternative Deep Learning based approach.
Raik is a visiting AI Bioinformatics Reseacher at Deepkapha.ai and a Bioinformatics Ph.D candidate. He is the author of the R-Bioconductor package called ‘Uniquorn’. He completed his Bachelor’s and Master’s in Bioinformatics from the Freie Universität Berlin and has been working in this field since then.