Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference.

Young AL, Marinescu RV, Oxtoby NP, Bocchetta M, Yong K, Firth NC, Cash DM, Thomas DL, Dick KM, Cardoso J, van Swieten J, Borroni B, Galimberti D, Masellis M, Tartaglia MC, Rowe JB, Graff C, Tagliavini F, Frisoni GB, Frisoni GB, Laforce R, Finger E, de Mendonça A, Sorbi S, Warren JD, Crutch S, Fox NC, Ourselin S, Schott JM, Rohrer JD, Alexander DC, Genetic FTD Initiative (GENFI), Alzheimer’s Disease Neuroimaging Initiative (ADNI). · 2018

Read Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference. by Young AL, Marinescu RV, Oxtoby NP, Bocchetta M, Yong K, Firth NC, Cash DM, Thomas DL, Dick KM, Cardoso J, van Swieten J, Borroni B, Galimberti D, Masellis M, Tartaglia MC, Rowe JB, Graff C, Tagliavini F, Frisoni GB, Frisoni GB, Laforce R, Finger E, de Mendonça A, Sorbi S, Warren JD, Crutch S, Fox NC, Ourselin S, Schott JM, Rohrer JD, Alexander DC, Genetic FTD Initiative (GENFI), Alzheimer’s Disease Neuroimaging Initiative (ADNI). (2018) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc6189176.txt.

Introduction

Young and co-authors present a methods article in Nature Communications on a machine-learning tool for untangling brain diseases. Neurodegenerative diseases vary widely between patients. Study groups mix different forms on different trajectories, which confuses research and drug trials. Standard studies capture each patient at just one time point.

The tool, called SuStaIn, infers both subtype and stage from such cross-sectional data. Applied to imaging in genetic frontotemporal dementia, it found subgroups with distinct patterns of brain shrinkage. It could even identify the genetic type from imaging alone. Work in Alzheimer's disease data showed the same skill. The method promises cleaner groups for trials and for tracking disease over time.

Catalog record

CREATORYoung AL, Marinescu RV, Oxtoby NP, Bocchetta M, Yong K, Firth NC, Cash DM, Thomas DL, Dick KM, Cardoso J, van Swieten J, Borroni B, Galimberti D, Masellis M, Tartaglia MC, Rowe JB, Graff C, Tagliavini F, Frisoni GB, Frisoni GB, Laforce R, Finger E, de Mendonça A, Sorbi S, Warren JD, Crutch S, Fox NC, Ourselin S, Schott JM, Rohrer JD, Alexander DC, Genetic FTD Initiative (GENFI), Alzheimer’s Disease Neuroimaging Initiative (ADNI).
YEAR2018
COLLECTIONPAPERS
FORMATTXT
FILE SIZE83 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC6189176

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference By Alexandra L Young, Razvan V Marinescu, Neil P Oxtoby, Martina Bocchetta, Keir Yong, Nicholas C Firth, David M Cash, David L Thomas, Katrina M Dick, Jorge Cardoso, John van Swieten, Barbara Borroni Published in: Nature Communications DOI: 10.1038/s41467-018-05892-0 ABSTRACT 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. INTRODUCTION Neurodegenerative disorders, such as frontotemporal dementia (FTD) and Alzheimer’s disease (AD), are biologically heterogeneous, producing high variance in in vivo disease biomarkers, such as volumetric measurements from imaging, protein measurements from lumbar puncture or behavioural measurements from psychometrics, which reduces their utility in disease studies and management. Key contributors to this heterogeneity are that individuals belong to a range of disease subtypes (giving rise to phenotypic heterogeneity) and are at different stages of a dynamic disease process (producing temporal heterogeneity). Previous studies aiming to explain biomarker variance typically focus on a single aspect of this heterogeneity: phenotypic heterogeneity at a coarse, typically late, disease stage or temporal heterogeneity in a broad population. However, the inability to disentangle the range of subtypes from the development and progression of each over time limits the biological insight these techniques can provide, as well as their utility for patient stratification. Constructing a comprehensive picture separating phenotypic and temporal heterogeneity, i.e. identifying distinct subtypes and characterising the development and progression of each, remains a major current challenge. However, such a picture would provide insights into underlying disease mechanisms, and enable accurate fine-grained patient stratification and prognostication, facilitating precision medicine in clinical trials and healthcare. Both FTD and AD exhibit substantial pathologic, genetic and clinical heterogeneity. In FTD a large proportion of cases (around a third) are inherited on an autosomal-dominant basis, with mutations in progranulin (GRN), microtubule-associated protein tau (MAPT) and chromosome 9 open reading frame 72 (C9orf72) being the most common causes. Of the major genetic groups, GRN mutations are associated with TDP-43 type A pathology, MAPT mutations with tau inclusions, and expansions in C9orf72 with type A or type B TDP-43 pathology1. AD instead has a single pathological characterisation: the presence of both amyloid plaques and neurofibrillary tangles, and the proportion of autosomal-dominant cases is much smaller, accounting for between 1 and 6% of cases2. The pathological heterogeneity observed in AD consists of variation in the distribution of neurofibrillary tangles, with 25% of patients having an atypical distribution of neurofibrillary tangles (described as hippocampal-sparing or limbic-predominant) on autopsy at the time of death3. Both FTD and AD exhibit a diverse range of clinical syndromes. FTD has both behavioural and language presentations, and in genetic FTD the clinical syndromes can further include atypical parkinsonism and amyotrophic lateral sclerosis. In AD, the major clinical syndrome is broadly divided into amnestic and rarer non-amnestic variants, with non-amnestic variants including language variant AD, logopenic progressive aphasia, visuoperceptive variant AD, posterior cortical atrophy and frontal variant AD4. Previous studies of neurodegenerative disease heterogeneity have focussed on either temporal heterogeneity (i.e. subjects appear different at different disease stages) or phenotypic heterogeneity (i.e. distinct groups of subjects appear different even at the same disease stage), but rarely both. We refer to these two approaches as stages-only models, which account for temporal heterogeneity but not phenotypic heterogeneity, and subtypes-only models, which account for phenotypic heterogeneity but not temporal heterogeneity. Stages-only models arise for example from regression against disease stage5,6, and data-driven disease progression modelling7–15. Although such models have enabled deeper understanding of the temporal progression of a range of conditions, the inherent assumption that all individuals have a single phenotype, i.e. follow approximately the same trajectory, is a key limitation. At best, this limits the biological insight and the accuracy of stratification they can provide, but potentially could also lead to erroneous conclusions. Subtypes-only models use, for example, clustering (e.g. refs. 16–23) to identify distinct groups, or group individuals using information independent of the model, such as genetics (e.g. ref. 24) or post-mortem examination (e.g. refs. 25–28) for models based on in vivo imaging. With typical subtypes-only models, the limitation is the inherent assumption that all subjects are at a common disease stage so that the cohort has no temporal heterogeneity. This requires a priori staging and selection of individuals, which is typically crude in practice leaving models that are not specific to subtype differences. Models of both disease subtype and stage heterogeneity have been constructed previously for the small proportion of neurodegenerative diseases that are inherited on an autosomal-dominant basis. For example, Rohrer et al.29 investigate temporal heterogeneity within genetic groups by regressing imaging markers against an estimated age of onset (from family history). However, such studies lack the ability to identify within-genotype phenotypes, and the temporal resolution of the recovered genotype progression patterns is limited by inaccuracy of the a priori staging. This paper presents Subtype and Stage Inference (SuStaIn): a computational technique that disentangles temporal and phenotypic heterogeneity to identify population subgroups with common patterns of disease progression. We demonstrate SuStaIn using structural magnetic resonance imaging (MRI) data sets from cohorts of genetic FTD and AD patients. In each case, SuStaIn provides a data-driven taxonomy (set of subtypes and stages), as well as detailed pictures of the progression of neurodegeneration within each of the data-driven subgroups. From the genetic FTD data set, SuStaIn identifies subtypes from imaging alone that map closely onto the genotypes and reconstructs patterns of neurodegeneration that reflect analysis of the individual genetic groups. This provides a validation of SuStaIn’s ability to identify subgroups with distinct temporal progression patterns, as the different genotypes are known to have distinct patterns of neurodegeneration visible as brain atrophy in MRI29. However, SuStaIn further uncovers two distinct within-genotype phenotypes for carriers of a mutation in the C9orf72 gene, while finding the MAPT and GRN mutation groups are more homogeneous. In AD, SuStaIn identifies three distinct subtypes and reconstructs their previously unseen temporal progression. In both neurodegenerative diseases, we demonstrate strong assignment of individuals to the SuStaIn subtypes, which is in contrast to subtypes-only models in the literature (e.g. ref. 23). Even at very early stages, at least a proportion of individuals show strong alignment with particular subtypes, which highlights the potential utility in precision medicine. In AD, we show that SuStaIn subtype and stage enhance the ability to predict conversion between diagnostic categories substantially beyond subtypes-only or stages-only models. RESULTS Subtype and stage inference

Read the complete text