A robust unsupervised clustering approach for high-dimensional biological imaging data reveals shared drug-induced morphological signatures

Abstract

AbstractHigh-throughput analysis methods have emerged as central technologies to accelerate discovery by accelerating generation of large-scale data. Analysis and interpretation of these datasets remains challenging due to limitations in computational approaches for dimensionality reduction. Here, we present UnTANGLeD, a versatile computational pipeline that prioritises biologically robust and meaningful information to guide actionable strategies from input screening data which we demonstrate using results from image-based drug screening. By providing a robust framework for analysing high dimensional biological data, UnTANGLeD offers a powerful tool for analysis of theoretically any data type from any screening platform.