omicSynth: An open multi-omic community resource for identifying druggable targets across neurodegenerative diseases.

Treatments for neurodegenerative disorders remain rare, but recent FDA approvals, such as lecanemab and aducanumab for Alzheimer disease (MIM: 607822), highlight the importance of the underlying biological mechanisms in driving discovery and creating disease modifying therapies. The global population is aging, driving an urgent need for therapeutics that stop disease progression and eliminate symptoms. In this study, we create an open framework and resource for evidence-based identification of therapeutic targets for neurodegenerative disease. We use summary-data-based Mendelian randomization to identify genetic targets for drug discovery and repurposing. In parallel, we provide mechanistic insights into disease processes and potential network-level consequences of gene-based therapeutics. We identify 116 Alzheimer disease, 3 amyotrophic lateral sclerosis (MIM: 105400), 5 Lewy body dementia (MIM: 127750), 46 Parkinson disease (MIM: 605909), and 9 progressive supranuclear palsy (MIM: 601104) target genes passing multiple test corrections (pSMR_multi < 2.95 × 10-6 and pHEIDI > 0.01). We created a therapeutic scheme to classify our identified target genes into strata based on druggability and approved therapeutics, classifying 41 novel targets, 3 known targets, and 115 difficult targets (of these, 69.8% are expressed in the disease-relevant cell type from single-nucleus experiments). Our novel class of genes provides a springboard for new opportunities in drug discovery, development, and repurposing in the pre-competitive space. In addition, looking at drug-gene interaction networks, we identify previous trials that may require further follow-up such as riluzole in Alzheimer disease. We also provide a user-friendly web platform to help users explore potential therapeutic targets for neurodegenerative diseases, decreasing activation energy for the community.

Genetic risk factor clustering within and across neurodegenerative diseases.

Overlapping symptoms and co-pathologies are common in closely related neurodegenerative diseases (NDDs). Investigating genetic risk variants across these NDDs can give further insight into disease manifestations. In this study we have leveraged genome-wide single nucleotide polymorphisms and genome-wide association study summary statistics to cluster patients based on their genetic status across identified risk variants for five NDDs (Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis, Lewy body dementia and frontotemporal dementia). The multi-disease and disease-specific clustering results presented here provide evidence that NDDs have more overlapping genetic aetiology than previously expected and how neurodegeneration should be viewed as a spectrum of symptomology. These clustering analyses also show potential subsets of patients with these diseases that are significantly depleted for any known common genetic risk factors suggesting environmental or other factors at work. Establishing that NDDs with overlapping pathologies share genetic risk loci, future research into how these variants might have different effects on downstream protein expression, pathology and NDD manifestation in general is important for refining and treating NDDs.

PhenoExam: gene set analyses through integration of different phenotype databases.

BACKGROUND: Gene set enrichment analysis (detecting phenotypic terms that emerge as significant in a set of genes) plays an important role in bioinformatics focused on diseases of genetic basis. To facilitate phenotype-oriented gene set analysis, we developed PhenoExam, a freely available R package for tool developers and a web interface for users, which performs: (1) phenotype and disease enrichment analysis on a gene set; (2) measures statistically significant phenotype similarities between gene sets and (3) detects significant differential phenotypes or disease terms across different databases. RESULTS: PhenoExam generates sensitive and accurate phenotype enrichment analyses. It is also effective in segregating gene sets or Mendelian diseases with very similar phenotypes. We tested the tool with two similar diseases (Parkinson and dystonia), to show phenotype-level similarities but also potentially interesting differences. Moreover, we used PhenoExam to validate computationally predicted new genes potentially associated with epilepsy. CONCLUSIONS: We developed PhenoExam, a freely available R package and Web application, which performs phenotype enrichment and disease enrichment analysis on gene set G, measures statistically significant phenotype similarities between pairs of gene sets G and G' and detects statistically significant exclusive phenotypes or disease terms, across different databases. We proved with simulations and real cases that it is useful to distinguish between gene sets or diseases with very similar phenotypes. Github R package URL is https://github.com/alexcis95/PhenoExam . Shiny App URL is https://alejandrocisterna.shinyapps.io/phenoexamweb/ .

Differences in the Presentation and Progression of Parkinson's Disease by Sex.

BACKGROUND: Previous studies reported various symptoms of Parkinson's disease (PD) associated with sex. Some were conflicting or confirmed in only one study. OBJECTIVES: We examined sex associations to PD phenotypes cross-sectionally and longitudinally in large-scale data. METHODS: We tested 40 clinical phenotypes, using longitudinal, clinic-based patient cohorts, consisting of 5946 patients, with a median follow-up of 3.1 years. For continuous outcomes, we used linear regressions at baseline to test sex-associated differences in presentation, and linear mixed-effects models to test sex-associated differences in progression. For binomial outcomes, we used logistic regression models at baseline and Cox regression models for survival analyses. We adjusted for age, disease duration, and medication use. In the secondary analyses, data from 17 719 PD patients and 7588 non-PD participants from an online-only, self-assessment PD cohort were cross-sectionally evaluated to determine whether the sex-associated differences identified in the primary analyses were consistent and unique to PD. RESULTS: Female PD patients had a higher risk of developing dyskinesia early during the follow-up period, with a slower progression in activities of daily living difficulties, and a lower risk of developing cognitive impairments compared with male patients. The findings in the longitudinal, clinic-based cohorts were mostly consistent with the results of the online-only cohort. CONCLUSIONS: We observed sex-associated contributions to PD heterogeneity. These results highlight the necessity of future research to determine the underlying mechanisms and importance of personalized clinical management. © 2020 International Parkinson and Movement Disorder Society.

Genetic variability and potential effects on clinical trial outcomes: perspectives in Parkinson's disease.

BACKGROUND: Classical randomisation of clinical trial patients creates a source of genetic variance that may be contributing to the high failure rate seen in neurodegenerative disease trials. Our objective was to quantify genetic difference between randomised trial arms and determine how imbalance can affect trial outcomes. METHODS: 5851 patients with Parkinson's disease of European ancestry data and two simulated virtual cohorts based on public data were used. Data were resampled at different sizes for 1000 iterations and randomly assigned to the two arms of a simulated trial. False-negative and false-positive rates were estimated using simulated clinical trials, and per cent difference in genetic risk score (GRS) and allele frequency was calculated to quantify variance between arms. RESULTS: 5851 patients with Parkinson's disease (mean (SD) age, 61.02 (12.61) years; 2095 women (35.81%)) as well as simulated patients from virtually created cohorts were used in the study. Approximately 90% of the iterations had at least one statistically significant difference in individual risk SNPs between each trial arm. Approximately 5%-6% of iterations had a statistically significant difference between trial arms in mean GRS. For significant iterations, the average per cent difference for mean GRS between trial arms was 130.87%, 95% CI 120.89 to 140.85 (n=200). Glucocerebrocidase (GBA) gene-only simulations see an average 18.86%, 95% CI 18.01 to 19.71 difference in GRS scores between trial arms (n=50). When adding a drug effect of -0.5 points in MDS-UPDRS per year at n=50, 33.9% of trials resulted in false negatives. CONCLUSIONS: Our data support the hypothesis that within genetically unmatched clinical trials, genetic heterogeneity could confound true therapeutic effects as expected. Clinical trials should undergo pretrial genetic adjustment or, at the minimum, post-trial adjustment and analysis for failed trials.

Penetrance of Parkinson's Disease in LRRK2 p.G2019S Carriers Is Modified by a Polygenic Risk Score.

BACKGROUND: Although the leucine-rich repeat kinase 2 p.G2019S mutation has been demonstrated to be a strong risk factor for PD, factors that contribute to penetrance among carriers, other than aging, have not been well identified. OBJECTIVES: To evaluate whether a cumulative genetic risk identified in the recent genome-wide study is associated with penetrance of PD among p.G2019S mutation carriers. METHODS: We included p.G2019S heterozygote carriers with European ancestry in three genetic cohorts in which the mutation carriers with and without PD were selectively recruited. We also included the carriers from two data sets: one from a case-control setting without selection of mutation carriers and the other from a population sampling. Associations between polygenic risk score constructed from 89 variants reported recently and PD were tested and meta-analyzed. We also explored the interaction of age and PRS. RESULTS: After excluding eight homozygotes, 833 p.G2019S heterozygote carriers (439 PD and 394 unaffected) were analyzed. Polygenic risk score was associated with a higher penetrance of PD (odds ratio: 1.34; 95% confidence interval: [1.09, 1.64] per +1 standard deviation; P = 0.005). In addition, associations with polygenic risk score and penetrance were stronger in the younger participants (main effect: odds ratio 1.28 [1.04, 1.58] per +1 standard deviation; P = 0.022; interaction effect: odds ratio 0.78 [0.64, 0.94] per +1 standard deviation and + 10 years of age; P = 0.008). CONCLUSIONS: Our results suggest that there is a genetic contribution for penetrance of PD among p.G2019S carriers. These results have important etiological consequences and potential impact on the selection of subjects for clinical trials. © 2020 International Parkinson and Movement Disorder Society.

The Parkinson's Disease Genome-Wide Association Study Locus Browser.

BACKGROUND: Parkinson's disease (PD) is a neurodegenerative disease with an often complex component identifiable by genome-wide association studies. The most recent large-scale PD genome-wide association studies have identified more than 90 independent risk variants for PD risk and progression across more than 80 genomic regions. One major challenge in current genomics is the identification of the causal gene(s) and variant(s) at each genome-wide association study locus. The objective of the current study was to create a tool that would display data for relevant PD risk loci and provide guidance with the prioritization of causal genes and potential mechanisms at each locus. METHODS: We included all significant genome-wide signals from multiple recent PD genome-wide association studies including themost recent PD risk genome-wide association study, age-at-onset genome-wide association study, progression genome-wide association study, and Asian population PD risk genome-wide association study. We gathered data for all genes 1 Mb up and downstream of each variant to allow users to assess which gene(s) are most associated with the variant of interest based on a set of self-ranked criteria. Multiple databases were queried for each gene to collect additional causal data. RESULTS: We created a PD genome-wide association study browser tool (https://pdgenetics.shinyapps.io/GWASBrowser/) to assist the PD research community with the prioritization of genes for follow-up functional studies to identify potential therapeutic targets. CONCLUSIONS: Our PD genome-wide association study browser tool provides users with a useful method of identifying potential causal genes at all known PD risk loci from large-scale PD genome-wide association studies. We plan to update this tool with new relevant data as sample sizes increase and new PD risk loci are discovered. © 2020 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society. This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Genomewide association study of Parkinson's disease clinical biomarkers in 12 longitudinal patients' cohorts.

BACKGROUND: Several reports have identified different patterns of Parkinson's disease progression in individuals carrying missense variants in GBA or LRRK2 genes. The overall contribution of genetic factors to the severity and progression of Parkinson's disease, however, has not been well studied. OBJECTIVES: To test the association between genetic variants and the clinical features of Parkinson's disease on a genomewide scale. METHODS: We accumulated individual data from 12 longitudinal cohorts in a total of 4093 patients with 22,307 observations for a median of 3.81 years. Genomewide associations were evaluated for 25 cross-sectional and longitudinal phenotypes. Specific variants of interest, including 90 recently identified disease-risk variants, were also investigated post hoc for candidate associations with these phenotypes. RESULTS: Two variants were genomewide significant. Rs382940(T>A), within the intron of SLC44A1, was associated with reaching Hoehn and Yahr stage 3 or higher faster (hazard ratio 2.04 [1.58-2.62]; P value = 3.46E-8). Rs61863020(G>A), an intergenic variant and expression quantitative trait loci for α-2A adrenergic receptor, was associated with a lower prevalence of insomnia at baseline (odds ratio 0.63 [0.52-0.75]; P value = 4.74E-8). In the targeted analysis, we found 9 associations between known Parkinson's risk variants and more severe motor/cognitive symptoms. Also, we replicated previous reports of GBA coding variants (rs2230288: p.E365K; rs75548401: p.T408M) being associated with greater motor and cognitive decline over time, and an APOE E4 tagging variant (rs429358) being associated with greater cognitive deficits in patients. CONCLUSIONS: We identified novel genetic factors associated with heterogeneity of Parkinson's disease. The results can be used for validation or hypothesis tests regarding Parkinson's disease. © 2019 International Parkinson and Movement Disorder Society.

Big Data: Astronomical or Genomical?

Genomics is a Big Data science and is going to get much bigger, very soon, but it is not known whether the needs of genomics will exceed other Big Data domains. Projecting to the year 2025, we compared genomics with three other major generators of Big Data: astronomy, YouTube, and Twitter. Our estimates show that genomics is a "four-headed beast"--it is either on par with or the most demanding of the domains analyzed here in terms of data acquisition, storage, distribution, and analysis. We discuss aspects of new technologies that will need to be developed to rise up and meet the computational challenges that genomics poses for the near future. Now is the time for concerted, community-wide planning for the "genomical" challenges of the next decade.

Federated Learning for multi-omics: a performance evaluation in Parkinson's disease.

While machine learning (ML) research has recently grown more in popularity, its application in the omics domain is constrained by access to sufficiently large, high-quality datasets needed to train ML models. Federated Learning (FL) represents an opportunity to enable collaborative curation of such datasets among participating institutions. We compare the simulated performance of several models trained using FL against classically trained ML models on the task of multi-omics Parkinson's Disease prediction. We find that FL model performance tracks centrally trained ML models, where the most performant FL model achieves an AUC-PR of 0.876 ± 0.009, 0.014 ± 0.003 less than its centrally trained variation. We also determine that the dispersion of samples within a federation plays a meaningful role in model performance. Our study implements several open source FL frameworks and aims to highlight some of the challenges and opportunities when applying these collaborative methods in multi-omics studies.

Federated learning for multi-omics: A performance evaluation in Parkinson's disease.

While machine learning (ML) research has recently grown more in popularity, its application in the omics domain is constrained by access to sufficiently large, high-quality datasets needed to train ML models. Federated learning (FL) represents an opportunity to enable collaborative curation of such datasets among participating institutions. We compare the simulated performance of several models trained using FL against classically trained ML models on the task of multi-omics Parkinson's disease prediction. We find that FL model performance tracks centrally trained ML models, where the most performant FL model achieves an AUC-PR of 0.876 ± 0.009, 0.014 ± 0.003 less than its centrally trained variation. We also determine that the dispersion of samples within a federation plays a meaningful role in model performance. Our study implements several open-source FL frameworks and aims to highlight some of the challenges and opportunities when applying these collaborative methods in multi-omics studies.

GenoTools: An Open-Source Python Package for Efficient Genotype Data Quality Control and Analysis.

GenoTools, a Python package, streamlines population genetics research by integrating ancestry estimation, quality control (QC), and genome-wide association studies (GWAS) capabilities into efficient pipelines. By tracking samples, variants, and quality-specific measures throughout fully customizable pipelines, users can easily manage genetics data for large and small studies. GenoTools' "Ancestry" module renders highly accurate predictions, allowing for high-quality ancestry-specific studies, and enables custom ancestry model training and serialization, specified to the user's genotyping or sequencing platform. As the genotype processing engine that powers several large initiatives including the NIH's Center for Alzheimer's and Related Dementias (CARD) and the Global Parkinson's Genetics Program (GP2). GenoTools was used to process and analyze the UK Biobank and major Alzheimer's Disease (AD) and Parkinson's Disease (PD) datasets with over 400,000 genotypes from arrays and 5000 sequences and has led to novel discoveries in diverse populations. It has provided replicable ancestry predictions, implemented rigorous QC, and conducted genetic ancestry-specific GWAS to identify systematic errors or biases through a single command. GenoTools is a customizable tool that enables users to efficiently analyze and scale genotype data with reproducible and scalable ancestry, QC, and GWAS pipelines.

omicSynth: an Open Multi-omic Community Resource for Identifying Druggable Targets across Neurodegenerative Diseases.

Treatments for neurodegenerative disorders remain rare, although recent FDA approvals, such as Lecanemab and Aducanumab for Alzheimer's Disease, highlight the importance of the underlying biological mechanisms in driving discovery and creating disease modifying therapies. The global population is aging, driving an urgent need for therapeutics that stop disease progression and eliminate symptoms. In this study, we create an open framework and resource for evidence-based identification of therapeutic targets for neurodegenerative disease. We use Summary-data-based Mendelian Randomization to identify genetic targets for drug discovery and repurposing. In parallel, we provide mechanistic insights into disease processes and potential network-level consequences of gene-based therapeutics. We identify 116 Alzheimer's disease, 3 amyotrophic lateral sclerosis, 5 Lewy body dementia, 46 Parkinson's disease, and 9 Progressive supranuclear palsy target genes passing multiple test corrections (pSMR_multi < 2.95×10-6 and pHEIDI > 0.01). We created a therapeutic scheme to classify our identified target genes into strata based on druggability and approved therapeutics - classifying 41 novel targets, 3 known targets, and 115 difficult targets (of these 69.8% are expressed in the disease relevant cell type from single nucleus experiments). Our novel class of genes provides a springboard for new opportunities in drug discovery, development and repurposing in the pre-competitive space. In addition, looking at drug-gene interaction networks, we identify previous trials that may require further follow-up such as Riluzole in AD. We also provide a user-friendly web platform to help users explore potential therapeutic targets for neurodegenerative diseases, decreasing activation energy for the community [https://nih-card-ndd-smr-home-syboky.streamlit.app/].

Virus exposure and neurodegenerative disease risk across national biobanks.

With recent findings connecting the Epstein-Barr virus to an increased risk of multiple sclerosis and growing concerns regarding the neurological impact of the coronavirus pandemic, we examined potential links between viral exposures and neurodegenerative disease risk. Using time series data from FinnGen for discovery and cross-sectional data from the UK Biobank for replication, we identified 45 viral exposures significantly associated with increased risk of neurodegenerative disease and replicated 22 of these associations. The largest effect association was between viral encephalitis exposure and Alzheimer's disease. Influenza with pneumonia was significantly associated with five of the six neurodegenerative diseases studied. We also replicated the Epstein-Barr/multiple sclerosis association. Some of these exposures were associated with an increased risk of neurodegeneration up to 15 years after infection. As vaccines are currently available for some of the associated viruses, vaccination may be a way to reduce some risk of neurodegenerative disease.

Parallel CRISPR-Cas9 screens identify mechanisms of PLIN2 and lipid droplet regulation.

Despite the key roles of perilipin-2 (PLIN2) in governing lipid droplet (LD) metabolism, the mechanisms that regulate PLIN2 levels remain incompletely understood. Here, we leverage a set of genome-edited human PLIN2 reporter cell lines in a series of CRISPR-Cas9 loss-of-function screens, identifying genetic modifiers that influence PLIN2 expression and post-translational stability under different metabolic conditions and in different cell types. These regulators include canonical genes that control lipid metabolism as well as genes involved in ubiquitination, transcription, and mitochondrial function. We further demonstrate a role for the E3 ligase MARCH6 in regulating triacylglycerol biosynthesis, thereby influencing LD abundance and PLIN2 stability. Finally, our CRISPR screens and several published screens provide the foundation for CRISPRlipid (http://crisprlipid.org), an online data commons for lipid-related functional genomics data. Our study identifies mechanisms of PLIN2 and LD regulation and provides an extensive resource for the exploration of LD biology and lipid metabolism.

Application of Aligned-UMAP to longitudinal biomedical studies.

High-dimensional data analysis starts with projecting the data to low dimensions to visualize and understand the underlying data structure. Several methods have been developed for dimensionality reduction, but they are limited to cross-sectional datasets. The recently proposed Aligned-UMAP, an extension of the uniform manifold approximation and projection (UMAP) algorithm, can visualize high-dimensional longitudinal datasets. We demonstrated its utility for researchers to identify exciting patterns and trajectories within enormous datasets in biological sciences. We found that the algorithm parameters also play a crucial role and must be tuned carefully to utilize the algorithm's potential fully. We also discussed key points to remember and directions for future extensions of Aligned-UMAP. Further, we made our code open source to enhance the reproducibility and applicability of our work. We believe our benchmarking study becomes more important as more and more high-dimensional longitudinal data in biomedical research become available.

A fully automated FAIMS-DIA mass spectrometry-based proteomic pipeline.

Here, we present a standardized, "off-the-shelf" proteomics pipeline working in a single 96-well plate to achieve deep coverage of cellular proteomes with high throughput and scalability. This integrated pipeline streamlines a fully automated sample preparation platform, a data-independent acquisition (DIA) coupled with high-field asymmetric waveform ion mobility spectrometer (FAIMS) interface, and an optimized library-free DIA database search strategy. Our systematic evaluation of FAIMS-DIA showing single compensation voltage (CV) at -35 V not only yields the deepest proteome coverage but also best correlates with DIA without FAIMS. Our in-depth comparison of direct-DIA database search engines shows that Spectronaut outperforms others, providing the highest quantifiable proteins. Next, we apply three common DIA strategies in characterizing human induced pluripotent stem cell (iPSC)-derived neurons and show single-shot mass spectrometry (MS) using single-CV (-35 V)-FAIMS-DIA results in >9,000 quantifiable proteins with <10% missing values, as well as superior reproducibility and accuracy compared with other existing DIA methods.

Sleep disturbances as risk factors for neurodegeneration later in life.

The relationship between sleep disorders and neurodegeneration is complex and multi-faceted. Using over one million electronic health records (EHRs) from Wales, UK, and Finland, we mined biobank data to identify the relationships between sleep disorders and the subsequent manifestation of neurodegenerative diseases (NDDs) later in life. We then examined how these sleep disorders' severity impacts neurodegeneration risk. Additionally, we investigated how sleep attributed risk may compensate for the lack of genetic risk factors (i.e. a lower polygenic risk score) in NDD manifestation. We found that sleep disorders such as sleep apnea were associated with the risk of Alzheimer's disease (AD), amyotrophic lateral sclerosis, dementia, Parkinson's disease (PD), and vascular dementia in three national scale biobanks, with hazard ratios (HRs) ranging from 1.31 for PD to 5.11 for dementia. These sleep disorders imparted significant risk up to 15 years before the onset of an NDD. Cumulative number of sleep disorders in the EHRs were associated with a higher risk of neurodegeneration for dementia and vascular dementia. Sleep related risk factors were independent of genetic risk for Alzheimer's and Parkinson's, potentially compensating for low genetic risk in overall disease etiology. There is a significant multiplicative interaction regarding the combined risk of sleep disorders and Parkinson's disease. Poor sleep hygiene and sleep apnea are relatively modifiable risk factors with several treatment options, including CPAP and surgery, that could potentially reduce the risk of neurodegeneration. This is particularly interesting in how sleep related risk factors are significantly and independently enriched in manifesting NDD patients with low levels of genetic risk factors for these diseases.

NeuroBooster Array: A Genome-Wide Genotyping Platform to Study Neurological Disorders Across Diverse Populations.

Genome-wide genotyping platforms have the capacity to capture genetic variation across different populations, but there have been disparities in the representation of population-dependent genetic diversity. The motivation for pursuing this endeavor was to create a comprehensive genome-wide array capable of encompassing a wide range of neuro-specific content for the Global Parkinson's Genetics Program (GP2) and the Center for Alzheimer's and Related Dementias (CARD). CARD aims to increase diversity in genetic studies, using this array as a tool to foster inclusivity. GP2 is the first supported resource project of the Aligning Science Across Parkinson's (ASAP) initiative that aims to support a collaborative global effort aimed at significantly accelerating the discovery of genetic factors contributing to Parkinson's disease and atypical parkinsonism by generating genome-wide data for over 200,000 individuals in a multi-ancestry context. Here, we present the Illumina NeuroBooster array (NBA), a novel, high-throughput and cost-effective custom-designed content platform to screen for genetic variation in neurological disorders across diverse populations. The NBA contains a backbone of 1,914,934 variants (Infinium Global Diversity Array) complemented with custom content of 95,273 variants implicated in over 70 neurological conditions or traits with potential neurological complications. Furthermore, the platform includes over 10,000 tagging variants to facilitate imputation and analyses of neurodegenerative disease-related GWAS loci across diverse populations. The NBA can identify low frequency variants and accurately impute over 15 million common variants from the latest release of the TOPMed Imputation Server as of August 2023 (reference of over 300 million variants and 90,000 participants). We envisage this valuable tool will standardize genetic studies in neurological disorders across different ancestral groups, allowing researchers to perform genetic research inclusively and at a global scale.

Identifying and predicting amyotrophic lateral sclerosis clinical subgroups: a population-based machine-learning study.

BACKGROUND: Amyotrophic lateral sclerosis (ALS) is known to represent a collection of overlapping syndromes. Various classification systems based on empirical observations have been proposed, but it is unclear to what extent they reflect ALS population substructures. We aimed to use machine-learning techniques to identify the number and nature of ALS subtypes to obtain a better understanding of this heterogeneity, enhance our understanding of the disease, and improve clinical care. METHODS: In this retrospective study, we applied unsupervised Uniform Manifold Approximation and Projection [UMAP]) modelling, semi-supervised (neural network UMAP) modelling, and supervised (ensemble learning based on LightGBM) modelling to a population-based discovery cohort of patients who were diagnosed with ALS while living in the Piedmont and Valle d'Aosta regions of Italy, for whom detailed clinical data, such as age at symptom onset, were available. We excluded patients with missing Revised ALS Functional Rating Scale (ALSFRS-R) feature values from the unsupervised and semi-supervised steps. We replicated our findings in an independent population-based cohort of patients who were diagnosed with ALS while living in the Emilia Romagna region of Italy. FINDINGS: Between Jan 1, 1995, and Dec 31, 2015, 2858 patients were entered in the discovery cohort. After excluding 497 (17%) patients with missing ALSFRS-R feature values, data for 42 clinical features across 2361 (83%) patients were available for the unsupervised and semi-supervised analysis. We found that semi-supervised machine learning produced the optimum clustering of the patients with ALS. These clusters roughly corresponded to the six clinical subtypes defined by the Chiò classification system (ie, bulbar, respiratory, flail arm, classical, pyramidal, and flail leg ALS). Between Jan 1, 2009, and March 1, 2018, 1097 patients were entered in the replication cohort. After excluding 108 (10%) patients with missing ALSFRS-R feature values, data for 42 clinical features across 989 patients were available for the unsupervised and semi-supervised analysis. All 1097 patients were included in the supervised analysis. The same clusters were identified in the replication cohort. By contrast, other ALS classification schemes, such as the El Escorial categories, Milano-Torino clinical staging, and King's clinical stages, did not adequately label the clusters. Supervised learning identified 11 clinical parameters that predicted ALS clinical subtypes with high accuracy (area under the curve 0·982 [95% CI 0·980-0·983]). INTERPRETATION: Our data-driven study provides insight into the ALS population substructure and confirms that the Chiò classification system successfully identifies ALS subtypes. Additional validation is required to determine the accuracy and clinical use of these algorithms in assigning clinical subtypes. Nevertheless, our algorithms offer a broad insight into the clinical heterogeneity of ALS and help to determine the actual subtypes of disease that exist within this fatal neurodegenerative syndrome. The systematic identification of ALS subtypes will improve clinical care and clinical trial design. FUNDING: US National Institute on Aging, US National Institutes of Health, Italian Ministry of Health, European Commission, University of Torino Rita Levi Montalcini Department of Neurosciences, Emilia Romagna Regional Health Authority, and Italian Ministry of Education, University, and Research. TRANSLATIONS: For the Italian and German translations of the abstract see Supplementary Materials section.