Multiscale PHATE Exploration of SARS-CoV-2 Data Reveals Multimodal Signatures of Disease

The biomedical community is producing increasingly high dimensional datasets, integrated from hundreds of patient samples, which current computational techniques struggle to explore. Here we present Multiscale PHATE, which learns abstracted biological features from data that can be directly predictive of disease. Our approach creates a tree of data granularities that can be cut at coarse levels for high level summarizations, as well as at fine levels for detailed representations on subsets. We apply Multiscale PHATE to study the immune response to COVID-19 in 54 million cells from 168 hospitalized patients. Our analysis identifies pathogenic cellular populations, CD16-hiCD66b-lo neutrophils and IFNγ+GranzymeB+ Th17 cells, and shows that cellular groupings discovered by Multiscale PHATE are directly predictive of disease outcome. We use Multiscale PHATE-derived features to construct two different manifolds of patients, one from abstracted flow cytometry features and another on patient clinical features, both associating immune subsets and clinical markers with outcome. Conflict of Interest: Dr. Krishnaswamy is on the scientific advisory board of KovaDx and AI Therapeutics. Dr. Iwasaki a member of the SAB for InProTher. Dr. Iwasaki is a co-founder of RIGImmune. Dr. Wilson is founder of Efference. Dr. Ko is a member of the expert panel of the Reckit Global Hygiene Institute. The remaining authors have no competing interests to declare. Ethical Approval: This study was approved by Yale Human Research Protection Program Institutional Review Boards (FWA00002571, protocol ID 2000027690). Informed consent was obtained from all enrolled patients and healthcare workers.