The heterogeneity of neurodegenerative diseases is a key confound to disease understanding and treatment development, as study cohorts typically include multiple phenotypes on distinct disease trajectories. Here we introduce a machine-learning technique—Subtype and Stage Inference (SuStaIn)—able to uncover data-driven disease phenotypes with distinct temporal progression patterns, from widely available cross-sectional patient studies. Results from imaging studies in two neurodegenerative diseases reveal subgroups and their distinct trajectories of regional neurodegeneration. In genetic frontotemporal dementia, SuStaIn identifies genotypes from imaging alone, validating its ability to identify subtypes; further the technique reveals within-genotype heterogeneity. In Alzheimer’s disease, SuStaIn uncovers three subtypes, uniquely characterising their temporal complexity. SuStaIn provides fine-grained patient stratification, which substantially enhances the ability to predict conversion between diagnostic categories over standard models that ignore subtype (p = 7.18 × 10−4) or temporal stage (p = 3.96 × 10−5). SuStaIn offers new promise for enabling disease subtype discovery and precision medicine.

See Stankoff and Louapre (doi:) for a scientific commentary on this article.Grey matter atrophy in multiple sclerosis affects certain areas preferentially. Eshaghi et al. use a data-driven computational model to predict the order in which regions atrophy, and use this sequence to stage patients. Atrophy begins in deep grey matter nuclei and posterior cortical regions, before spreading to other cortical areas.

SummaryThe heterogeneity of neurodegenerative diseases is a key confound to disease understanding and treatment development, as study cohorts typically include multiple phenotypes on distinct disease trajectories. Here we present a new machine learning technique – Subtype and Stage Inference (SuStaIn) – able to uncover data-driven disease phenotypes with distinct temporal progression patterns, from widely available crosssectional patient studies. Results from imaging studies in two neurodegenerative diseases reveal new subgroups and their distinct trajectories of regional neurodegeneration. In genetic frontotemporal dementia, SuStaIn identifies genotypes from imaging alone, validating its ability to identify subtypes, and characterises within-group heterogeneity for the first time. In Alzheimer’s disease, SuStaIn uncovers three subtypes, uniquely revealing their temporal complexity. SuStaIn provides fine-grained patient stratification, which substantially enhances the ability to predict conversion between diagnostic categories over standard models that ignore subtype (p=7.18×10--4) or temporal stage (p=3.96×10−5). SuStaIn thus offers new promise for enabling disease subtype discovery and precision medicine.

We describe a computational method, plane of best fit (PBF), to quantify and characterize the 3D character of molecules. This method is rapid and amenable to analysis of large diverse data sets. We compare PBF with alternative literature methods used to assess 3D character and apply the method to diverse data sets of fragment-like, drug-like, and natural product compound libraries. We show that exemplar fragment libraries underexploit the potential of 3D character in fragment-like chemical space and that drug-like molecules in the libraries examined are predominantly 2D in character.