Uncovering the heterogeneity and temporal complexity of neurodegenerative diseases with Subtype and Stage Inference - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation

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

Alexandra L Young et al. Nat Commun. .

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.

PubMed Disclaimer

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
Conceptual overview of SuStaIn. The Underlying model panel (a) considers a patient cohort to consist of an unknown set of disease subtypes. The input data (Input data panel, b), which can be entirely cross-sectional, contains snapshots of biomarker measurements from each subject with unknown subtype and unknown temporal stage. SuStaIn recovers the set of disease subtypes and their temporal progression (as shown in the Output panel, c) via simultaneous clustering and disease progression modelling. Given a new snapshot, SuStaIn can estimate the probability the subject belongs to each subtype and stage, by comparing the snapshot with the reconstruction (as shown in the Application panel, d). This figure depicts two hypothetical disease subtypes, labelled I and II, and the biomarkers are regional brain volumes, but SuStaIn is readily applicable to any scalar disease biomarker and any number of subtypes. The colour of each region indicates the amount of pathology in that region, ranging from white (no pathology) to red to magenta to blue (maximum pathology)
Fig. 2
Fig. 2
SuStaIn modelling of genetic frontotemporal dementia using GENFI data. a The progression pattern of each of four subtypes that SuStaIn identifies. Each progression pattern consists of a sequence of stages in which regional brain volumes in mutation carriers (symptomatic and presymptomatic) reach different z-scores relative to non-carriers. Intuitively (for a more precise description see Methods: Uncertainty Estimation), at each stage the colour in each region indicates the level of severity of volume loss: white is unaffected; red is mildly affected (z-score of 1); magenta is moderately affected (z-score of 2); and blue is severely affected (z-score of 3 or more). The circle labelled A indicates the asymmetry of the atrophy pattern (absolute value of the difference in volume between the left and right hemispheres divided by the total volume of the left and right hemispheres) at each stage for each subtype. CVS is the model cross-validation similarity (see Methods: Similarity between two progression patterns): the average similarity of the subtype progression patterns across cross-validation folds, which ranges from 0 (no similarity) to 1 (maximum similarity). f is the proportion of participants estimated to belong to each subtype. b The contribution of each genotype to each of the SuStaIn subtypes. This is calculated as the probability an individual has a particular genotype given that they belong to a particular subtype
Fig. 3
Fig. 3
SuStaIn modelling of sporadic Alzheimer’s disease using ADNI data. The rows show the progression pattern of the three subtypes identified by SuStaIn. Diagrams as in Fig. 2, but the z-scores are measured relative to amyloid-negative (cerebrospinal fluid Aβ1–42 > 192 pg per ml) cognitively normal subjects, i.e. cognitively normal subjects with no evidence of amyloid pathology on cerebrospinal fluid. The cerebellum was not included as a region in the Alzheimer’s disease analysis and so is shaded in dark grey
Fig. 4
Fig. 4
Reproducibility of the SuStaIn subtypes in Fig. 3. A largely independent Alzheimer’s disease data set (only 59 subjects are in both the 576 subject data set used to generate this figure and the 793 subject data set used in Fig. 3) consisting of those with regional brain volume measurements from 1.5T MRI scans, rather than 3T MRI scans, are shown. Diagrams are as in Fig. 3, with the rows showing the progression pattern of one of the subtypes identified by SuStaIn. SuStaIn modelling identifies three major subtypes: a typical, a cortical and a subcortical subtype, which are in good agreement with the three subtypes in Fig. 3, as well as an additional very small outlier group (only 4%) with a subtype we term parietal. This small subgroup may represent outliers with a posterior cortical atrophy phenotype
Fig. 5
Fig. 5
SuStaIn subtyping and staging of genetic frontotemporal dementia and Alzheimer’s disease. a, b The assignability of the disease subtypes estimated by SuStaIn for genetic frontotemporal dementia, and Alzheimer’s disease. Each scatter plot visualises the probability that each individual belongs to each of the SuStaIn subtypes estimated for a genetic frontotemporal dementia (as shown in Fig. 2a), and b. Alzheimer’s disease (as shown in Fig. 3). In the triangle scatter plots, each of the corners corresponds to a probability of 1 of belonging to that subtype, and 0 for the other subtypes; the centre point of the triangle corresponds to a probability of 1/3 of belonging to each subtype. c, d The probability subjects from each of the diagnostic groups belong to each of the SuStaIn stages for c genetic frontotemporal dementia and d Alzheimer’s disease. CN = cognitively normal; MCI = mild cognitive impairment; AD = Alzheimer’s disease
Fig. 6
Fig. 6
Subtypes-only model for GENFI; not accounting for disease stage heterogeneity. Brain diagrams as in Fig. 2, but here each diagram represents a different subtype, which we refer to as severe frontal, severe temporal and mild frontotemporal. There is no notion of disease stage in the subtypes-only model
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Rohrer JD, Rosen HJ. Neuroimaging in frontotemporal dementia. Int. Rev. Psychiatry. 2013;25:221–229. doi: 10.3109/09540261.2013.778822. - DOI - PubMed
    1. Bekris LM, Yu CE, Bird TD, Tsuang DW. Review article: genetics of Alzheimer disease. J. Geriatr. Psychiatry Neurol. 2010;23:213–227. doi: 10.1177/0891988710383571. - DOI - PMC - PubMed
    1. Murray ME, et al. Neuropathologically defined subtypes of Alzheimer’s disease with distinct clinical characteristics:a retrospective study. Lancet Neurol. 2011;10:785–796. doi: 10.1016/S1474-4422(11)70156-9. - DOI - PMC - PubMed
    1. Lam Benjamin, Masellis Mario, Freedman Morris, Stuss Donald T, Black Sandra E. Clinical, imaging, and pathological heterogeneity of the Alzheimer's disease syndrome. Alzheimer's Research & Therapy. 2013;5(1):1. doi: 10.1186/alzrt155. - DOI - PMC - PubMed
    1. Bateman RJ, et al. Clinical and biomarker changes in dominantly inherited Alzheimer’s disease. N. Engl. J. Med. 2012;367:795–804. doi: 10.1056/NEJMoa1202753. - DOI - PMC - PubMed

Publication types