Dataset Information

Machine learning of COVID-19 clinical data identifies population structures with therapeutic potential.

ABSTRACT: Clinical outcomes for patients with COVID-19 are heterogeneous and there is interest in defining subgroups for prognostic modeling and development of treatment algorithms. We obtained 28 demographic and laboratory variables in patients admitted to hospital with COVID-19. These comprised a training cohort (n = 6099) and two validation cohorts during the first and second waves of the pandemic (n = 996; n = 1011). Uniform manifold approximation and projection (UMAP) dimension reduction and Gaussian mixture model (GMM) analysis was used to define patient clusters. 29 clusters were defined in the training cohort and associated with markedly different mortality rates, which were predictive within confirmation datasets. Deconvolution of clinical features within clusters identified unexpected relationships between variables. Integration of large datasets using UMAP-assisted clustering can therefore identify patient subgroups with prognostic information and uncovers unexpected interactions between clinical variables. This application of machine learning represents a powerful approach for delineating disease pathogenesis and potential therapeutic interventions.

SUBMITTER: Greenwood D

PROVIDER: S-EPMC9153184 | biostudies-literature | 2022 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine learning of COVID-19 clinical data identifies population structures with therapeutic potential.

Greenwood David D Taverner Thomas T Adderley Nicola J NJ Price Malcolm James MJ Gokhale Krishna K Sainsbury Christopher C Gallier Suzy S Welch Carly C Sapey Elizabeth E Murray Duncan D Fanning Hilary H Ball Simon S Nirantharakumar Krishnarajah K Croft Wayne W Moss Paul P

iScience 20220531 7

Clinical outcomes for patients with COVID-19 are heterogeneous and there is interest in defining subgroups for prognostic modeling and development of treatment algorithms. We obtained 28 demographic and laboratory variables in patients admitted to hospital with COVID-19. These comprised a training cohort (n = 6099) and two validation cohorts during the first and second waves of the pandemic (n = 996; n = 1011). Uniform manifold approximation and projection (UMAP) dimension reduction and Gaussian ...[more]

PMID: 35665240

Similar Datasets

Project description:COVID-19, the disease caused by SARS-CoV-2, has been disrupting our lives for more than two years now. SARS-CoV-2 interacts with human proteins to pave its way into the human body, thereby wreaking havoc. Moreover, the mutating variants of the virus that take place in the SARS-CoV-2 genome are also a cause of concern among the masses. Thus, it is very important to understand human-spike protein-protein interactions (PPIs) in order to predict new PPIs and consequently propose drugs for the human proteins in order to fight the virus and its different mutated variants, with the mutations occurring in the spike protein. This fact motivated us to develop a complete pipeline where PPIs and drug-protein interactions can be predicted for human-SARS-CoV-2 interactions. In this regard, initially interacting data sets are collected from the literature, and noninteracting data sets are subsequently created for human-SARS-CoV-2 by considering only spike glycoprotein. On the other hand, for drug-protein interactions both interacting and noninteracting data sets are considered from DrugBank and ChEMBL databases. Thereafter, a model based on a sequence-based feature is used to code the protein sequences of human and spike proteins using the well-known Moran autocorrelation technique, while the drugs are coded using another well-known technique, viz., PaDEL descriptors, to predict new human-spike PPIs and eventually new drug-protein interactions for the top 20 predicted human proteins interacting with the original spike protein and its different mutated variants like Alpha, Beta, Delta, Gamma, and Omicron. Such predictions are carried out by random forest as it is found to perform better than other predictors, providing an accuracy of 90.53% for human-spike PPI and 96.15% for drug-protein interactions. Finally, 40 unique drugs like eicosapentaenoic acid, doxercalciferol, ciclesonide, dexamethasone, methylprednisolone, etc. are identified that target 32 human proteins like ACACA, DST, DYNC1H1, etc.

Project description:Next-generation T-cell-directed vaccines for COVID-19 aim to induce durable T-cell immunity against circulating and future hypermutated SARS-CoV-2 variants. Mass Spectrometry (MS)based immunopeptidomics holds promise for guiding vaccine design, but computational challenges impede the precise and unbiased identification of conserved T-cell epitopes crucial for vaccines against rapidly mutating viruses. We introduce a computational framework and analysis platform integrating a novel machine learning algorithm, immunopeptidomics, intra-host data, epitope immunogenicity, and geo-temporal CD8+ T-cell epitope conservation analyses. Central to our approach is MHCvalidator, a novel artificial neural network algorithm enhancing MS-based immunopeptidomics sensitivity by modeling antigen presentation and sequence features. MHCvalidator identified a novel nonconventional SARS-CoV-2 T-cell epitope presented by B7 supertype molecules, originating from a +1-frameshift in a truncated Spike (S) antigen, supported by ribo-seq data. Intra-host analysis of SARS-CoV-2 proteomes from ~100,000 COVID-19 patients revealed a prevalent S antigen truncation in ~51% of cases, exposing a rich source of frameshifted viral antigens. Our framework includes EpiTrack, a new computational pipeline tracking global mutational dynamics of MHCvalidator-identified SARS-CoV-2 CD8+ epitopes from vaccine BNT162b4. While most vaccine-encoded CD8+ epitopes exhibit global conservation from January 2020 to October 2023, a highly immunodominant A*01-associated epitope, especially in hospitalized patients, undergoes substantial mutations in Delta and Omicron variants. Our approach unveils unprecedented SARS-CoV-2 T-cell epitopes, elucidates novel antigenic features, and underscores mutational dynamics of vaccine-relevant epitopes. The analysis platform is applicable to any viruses, and underscores the need for continual vigilance in T-cell vaccine development against the evolving landscape of hypermutating SARS-CoV-2 variants.

Dataset Information

Machine learning of COVID-19 clinical data identifies population structures with therapeutic potential.

Publications

Machine learning of COVID-19 clinical data identifies population structures with therapeutic potential.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets