Functional immune mapping with deep-learning enabled phenomics applied to immunomodulatory and COVID-19 drug discovery

This article has been Reviewed by the following groups

Read the full article See related articles

Abstract

Development of accurate disease models and discovery of immune-modulating drugs is challenged by the immune system’s highly interconnected and context-dependent nature. Here we apply deep-learning-driven analysis of cellular morphology to develop a scalable “phenomics” platform and demonstrate its ability to identify dose-dependent, high-dimensional relationships among and between immunomodulators, toxins, pathogens, genetic perturbations, and small and large molecules at scale. High-throughput screening on this platform demonstrates rapid identification and triage of hits for TGF-β- and TNF-α-driven phenotypes. We deploy the platform to develop phenotypic models of active SARS-CoV-2 infection and of COVID-19-associated cytokine storm, surfacing compounds with demonstrated clinical benefit and identifying several new candidates for drug repurposing. The presented library of images, deep learning features, and compound screening data from immune profiling and COVID-19 screens serves as a deep resource for immune biology and cellular-model drug discovery with immediate impact on the COVID-19 pandemic.

Article activity feed

  1. SciScore for 10.1101/2020.08.02.233064: (What is this?)

    Please note, not all rigor criteria are appropriate for all manuscripts.

    Table 1: Rigor

    Institutional Review Board Statementnot detected.
    RandomizationTreatments were randomized across treatment plates with a 6-log range of immune stimuli (typically 0.001-100 ng/mL) at 6 replicates each with acoustic transfer (Echo 555, Labcyte) and incubated 37°C for 24 or (complete immune stimuli panel) or 48 h (for PBMC with pembrolizumab or nivolumab).
    Blindingnot detected.
    Power Analysisnot detected.
    Sex as a biological variablenot detected.
    Cell Line Authenticationnot detected.

    Table 2: Resources

    Antibodies
    SentencesResources
    Primary antibody solution was added to a 1:100 final dilution (Iba1 antibody Abcam cat #ab5076) incubated overnight at 4°C.
    Iba1
    suggested: (Abcam Cat# ab5076, RRID:AB_2224402)
    After overnight incubation with primary antibody cells were washed with PBS, and secondary antibody solution was added (AlexaFluor 488 donkey anti-goat IgG, 1:1000 final dilution.
    anti-goat IgG
    suggested: None
    2 Nucleocapsid staining: After staining and imaging to establish high dimensional phenotypes, plates were rinsed once with Wash Buffer (1xHBSS + 0.02% sodium azide) before incubating with primary antibody raised against SARS-CoV-2 nucleocapsid protein for 60 mins at RT (Sino Biological catno.
    SARS-CoV-2 nucleocapsid protein
    suggested: (Bioss Cat# bsm-41414M, RRID:AB_2848129)
    Experimental Models: Cell Lines
    SentencesResources
    Immune stimuli or virus were added 24 hours post-seeding (HUVEC, Macrophage, Fibroblast) or 1 h (PBMC).
    HUVEC
    suggested: KCB Cat# KCB 200648YJ, RRID:CVCL_2959)
    SARS-CoV-2 propagation and controls: The USA-WA1/2020 strain of SARS-CoV-2 was propagated in Vero cells.
    Vero
    suggested: CLS Cat# 605372/p622_VERO, RRID:CVCL_0059)
    Viral titers were determined through standard tissue culture infectious dose 50% (TCID50) methods, where cytopathic effect (CPE) on Vero 76 cells was measured by visual observation under a light microscope.
    Vero 76
    suggested: IZSLER Cat# BS CL 101, RRID:CVCL_0603)
    Software and Algorithms
    SentencesResources
    After 24 h incubation (or 96 hours post viral infection), plates were imaged using Image Express Micro Confocal High-Content Imaging System (
    Image Express
    suggested: None
    Images were analyzed with CellProfiler.
    CellProfiler
    suggested: None
    ALK5 biochemical assay: A 1536-well plate was pre-treated with compounds, at 13 concentrations (ranging from 0 to 100 ng/mL) with at least 2 replicates of each concentration, and a reaction mix containing 15 ng ALK5 (ThermoFisher), using Poly 4:1 as substrate was added to the plate.
    ThermoFisher
    suggested: (ThermoFisher; SL 8; Centrifuge, RRID:SCR_020809)

    Results from OddPub: We did not detect open data. We also did not detect open code. Researchers are encouraged to share open data when possible (see Nature blog).


    Results from LimitationRecognizer: We detected the following sentences addressing limitations in the study:
    This work and the recent success of dexamethasone in clinical trials for COVID-19 also identified a key limitation of our current phenomics approach: when studying a cell type in isolation, phenomics surfaces compounds that act via cell-autonomous mechanisms76. Compounds that intervene in multicellular processes might be revealed by development of co-culture models. Taken together, our results demonstrate that systems-level modeling and drug discovery is achievable using a single phenomics platform. First, this approach simplifies and extends the ability to work across many disease models rapidly because assay development work for any new model is minimized. Second, this work partially overcomes a historical limitation of phenotypic screening, predicting mechanism of action, by relating the high-dimensional phenoprint of hit compounds to those of reference molecules. Finally, we show the potential of this platform in optimizing NCE compounds through medicinal chemistry in a high-dimensional, target-agnostic manner. Unlike other high-dimensional approaches, the relatively inexpensive nature of these image-based assays allows them to be scaled to levels of throughput comparable to more traditional low-dimensional screening modalities. In the hopes that it will be valuable to others, we have made images and embeddings from HUVEC treated with the immune perturbant library, and from our COVID-19 primary screens (both infection and cytokine storm) available online (including raw im...

    Results from TrialIdentifier: No clinical trial numbers were referenced.


    Results from Barzooka: We did not find any issues relating to the usage of bar graphs.


    Results from JetFighter: We did not find any issues relating to colormaps.


    Results from rtransparent:
    • Thank you for including a conflict of interest statement. Authors are encouraged to include this statement when submitting to a journal.
    • No funding statement was detected.
    • No protocol registration statement was detected.

    About SciScore

    SciScore is an automated tool that is designed to assist expert reviewers by finding and presenting formulaic information scattered throughout a paper in a standard, easy to digest format. SciScore checks for the presence and correctness of RRIDs (research resource identifiers), and for rigor criteria such as sex and investigator blinding. For details on the theoretical underpinning of rigor criteria and the tools shown here, including references cited, please follow this link.