Inferring disease subtypes from clusters in explanation space

Scientific Reports, vol. 10, pp. 12900

Abstract

Identification of disease subtypes and corresponding biomarkers can substantially improve clinical diagnosis and treatment selection. Discovering these subtypes in noisy, high dimensional biomedical data is often impossible for humans and challenging for machines. We introduce a new approach to facilitate the discovery of disease subtypes: Instead of analyzing the original data, we train a diagnostic classifier (healthy vs. diseased) and extract instance-wise explanations for the classifier's decisions. The distribution of instances in the explanation space of our diagnostic classifier amplifies the different reasons for belonging to the same class-resulting in a representation that is uniquely useful for discovering latent subtypes. We compare our ability to recover subtypes via cluster analysis on model explanations to classical cluster analysis on the original data. In multiple datasets with known ground-truth subclasses, particularly on UK Biobank brain imaging data and transcriptome data from the Cancer Genome Atlas, we show that cluster analysis on model explanations substantially outperforms the classical approach. While we believe clustering in explanation space to be particularly valuable for inferring disease subtypes, the method is more general and applicable to any kind of sub-type identification.

Authors 5

1. Marc-André Schulz corresponding Aachen Department of Psychiatry, Psychotherapy and Psychosomatics

   RWTH Aachen University

   Affiliation as printed

   Department of Psychiatry, Psychotherapy and Psychosomatics, RWTH Aachen University, Aachen, Germany. marc.schulz@rwth-aachen.de

   QuantumBlack, London, UK. marc.schulz@rwth-aachen.de

   RWTH - Rheinisch-Westfälische Technische Hochschule Aachen University (RWTH Aachen Templergraben 55 52062 Aachen (Hausanschrift) 52056 Aachen (Postanschrift) - Germany)

2. Matt Chapman-Rounds

   University of Edinburgh

   Affiliation as printed

   QuantumBlack, London, UK

   School of Informatics, University of Edinburgh, Edinburgh, UK

   Edin. - University of Edinburgh (Old College South Bridge Edinburgh EH8 9YL - United Kingdom)

3. Manisha Verma

   Verizon Media (United States)

   Affiliation as printed

   QuantumBlack, London, UK

   Verizon Media, Boston, MA, USA

   QuantumBlack [London] (The Post Building, 100 Museum St, London WC1A 1PB, Royaume-Uni - United Kingdom)

4. Danilo Bzdok

   Montreal Neurological Institute and Hospital · Commissariat à l'Énergie Atomique et aux Énergies Alternatives · CEA Paris-Saclay - Etablissement de Saclay · Mila - Quebec Artificial Intelligence Institute · McGill University

   Affiliation as printed

   Department of Biomedical Engineering, McConnell Brain Imaging Centre (BIC), Montreal Neurological Institute (MNI), Faculty of Medicine, McGill University, Montreal, Canada

   Mila - Quebec Artificial Intelligence Institute, Montreal, Canada

   MILA - Montreal Institute for Learning Algorithms [Montréal] (6666 St-Urbain, #200, Montreal, QC, H2S 3H1, Canada - Canada)

   Montreal Neurological Hospital (3801 University Street Montreal, Quebec H3A 2B4 - Canada)

   PARIETAL - Modelling brain structure, function and variability based on high-field MRI data (Neurospin, CEA Saclay, Bâtiment 145, 91191 Gif-sur-Yvette Cedex - France)

5. Konstantinos Georgatzis
   Affiliation as printed

   QuantumBlack, London, UK

   QuantumBlack [London] (The Post Building, 100 Museum St, London WC1A 1PB, Royaume-Uni - United Kingdom)