Biology subjects

Bryson, B. D.

Publications and source records attributed to Bryson, B. D..

3 recordsLinked to original sources

bioRxiv2020

Learning with uncertainty for biological discovery and design

Machine learning that generates biological hypotheses has transformative potential, but most learning algorithms are susceptible to pathological failure when exploring regimes beyond the training data distribution. A solution is to quantify prediction uncertainty so that algorithms can gracefully handle novel phenomena that confound standard methods. Here, we demonstrate the broad utility of robust uncertainty prediction in biological discovery. By leveraging Gaussian process-based uncertainty prediction on modern pretrained features, we train a model on just 72 compounds to make predictions over a 10,833-compound library, identifying and experimentally validating compounds with nanomolar affinity for diverse kinases and whole-cell growth inhibition of Mycobacterium tuberculosis. We show how uncertainty facilitates a tight iterative loop between computation and experimentation, improves the generative design of novel biochemical structures, and generalizes across disparate biological domains. More broadly, our work demonstrates that uncertainty should play a key role in the increasing adoption of machine learning algorithms into the experimental lifecycle.

bioinformatics↗

bioRxiv2020

Consensus transcriptional states describe human mononuclear phagocyte diversity in the lung across health and disease

Monocytes, dendritic cells, and macrophages, commonly referred to as mononuclear phagocytes (MNPs), are innate immune cells capable of adopting diverse homeostatic and pathogenic phenotypes. Recent single-cell RNA-sequencing studies across many diseases in the lung have profiled this diversity transcriptionally, defining new cellular states and their association with disease. Despite these massive cellular profiling efforts, many studies have focused on defining myeloid dysfunction in specific diseases without identifying common pan-disease trends in the mononuclear phagocyte compartment within the lung. To address these gaps in our knowledge, we collate, process, and analyze 561,390 cellular transcriptomes from 12 studies of the human lung across multiple human diseases. We develop a computational framework to identify and compare dominant gene markers and gene expression programs and characterize MNP diversity in the lung, proposing a conserved dictionary of gene sets. Utilizing this reference, we efficiently identify disease-associated and rare MNP populations across multiple diseases and cohorts. Furthermore, we demonstrate the utility of this dictionary in characterizing a recently published dataset of bronchoalveolar lavage cells from COVID-19 patients and healthy controls which further reveal novel transcriptional shifts directly relatable to other diseases in the lung. These results underline conserved MNP transcriptional programs in lung disease, provide an immediate reference for characterizing the landscape of lung MNPs and establish a roadmap to dissecting MNP transcriptional complexity across tissues.

immunology↗

bioRxiv2020

Dissecting the common and compartment-specific features of COVID-19 severity in the lung and periphery with single-cell resolution

As the global COVID-19 pandemic continues to escalate, no effective treatment has yet been developed for the severe respiratory complications of this disease. This may be due in large part to the unclear immunopathological basis for the development of immune dysregulation and acute respiratory distress syndrome (ARDS) in severe and critical patients. Specifically, it remains unknown whether the immunological features of the disease that have been identified so far are compartment-specific responses or general features of COVID-19. Additionally, readily detectable biological markers correlated with strata of disease severity that could be used to triage patients and inform treatment options have not yet been identified. Here, we leveraged publicly available single-cell RNA sequencing data to elucidate the common and compartment-specific immunological features of clinically severe COVID-19. We identified a number of transcriptional programs that are altered across the spectrum of disease severity, few of which are common between the lung and peripheral immune environments. In the lung, comparing severe and moderate patients revealed severity-specific responses of enhanced interferon, A20/I{kappa}B, IL-2, and IL-6 pathway signatures along with broad signaling activity of IFNG, SPP1, CCL3, CCL8, and IL18 across cell types. These signatures contrasted with features unique to ARDS observed in the blood compartment, which included depletion of interferon and A20/I{kappa}B signatures and a lack of IL-6 response. The cell surface marker S1PR1 was strongly upregulated in patients diagnosed with ARDS compared to non-ARDS patients in {gamma}{delta} T cells of the blood compartment, and we nominate S1PR1 as a potential marker for immunophenotyping ARDS in COVID-19 patients using flow cytometry. HIGHLIGHTSO_LICOVID-19 disease severity is associated with a number of compositional shifts in the cellular makeup of the blood and lung environments. C_LIO_LITranscriptional data suggest differentially expressed cell surface proteins as markers for COVID-19 immunophenotyping from BALF and PBMC samples. C_LIO_LISeverity-specific features COVID-19 manifest at the pathway level, suggesting distinct changes to epithelia and differences between local and systemic immune dynamics. C_LIO_LIImmune-epithelial cellular communication analysis identifies ligands implicated in transcriptional regulation of proto-oncogenes in the lung epithelia of severe COVID-19 patients. C_LIO_LINetwork analysis suggests broadly-acting dysregulatory ligands in the pulmonary microenvironment as candidate therapeutic targets for the treatment of severe COVID-19. C_LI

immunology↗