SUPR
AI-Driven In Silico Screening of Anti-Cytokine Autoantibodies
Dnr:

NAISS 2025/22-54

Type:

NAISS Small Compute

Principal Investigator:

Rui Sun

Affiliation:

Karolinska Institutet

Start Date:

2025-02-13

End Date:

2026-03-01

Primary Classification:

30110: Immunology in the Medical Area

Webpage:

Allocation

Abstract

Project Overview Anti-cytokine autoantibodies (autoAbs) are present in healthy individuals and play a role in regulating excessive immune responses to cytokines. However, high level of neutralizing autoAbs against cytokines lead to impaired immune responses under pathological conditions. Notably, neutralizing autoAbs against type I interferons (IFNs) have been implicated in severe COVID-19 and other viral infections. Despite increasing recognition of their impact, there is a lack of systematic characterization of these autoAbs, particularly in terms of their gene expression profiles, immunogenetic features, and functional epitopes. This project leverages AI-driven in silico screening to identify and characterize anti-cytokine autoAbs, particularly those targeting type I IFNs, using computational approaches such as structural modeling, deep learning-based sequence analysis, and antigen-antibody interaction algorithm. We integrate AlphaFold-based structural modeling with single-cell RNA sequencing (scRNAseq) and fluorescence-activated cell sorting (FACS) to map the autoreactive B cell response. Objectives: 1. AI-Based AutoAb Identification: Utilize the Protein Data Bank (PDB) database and AlphaFold software for autoreactive B cell receptor (BCR) sequence prediction. 2. AlphaFold-Driven Structural Analysis: Apply molecular docking and AlphaFold-based structure prediction to assess autoAb binding sites. Characterize IFN-α2 epitope interactions and autoAb neutralization potential. 3. Integration with scRNAseq and FACS: The candidate clones screened by the AlphaFold will be further analyzed for their transcriptomic features and FACS phenotypes. 4. Experimental validation for the AlphaFold predicted clones: The autoAbs screened by the in silico approach will be produced as monoclonal antibodies. Further, we will analyze the biological functions of the produced autoAbs by ELISA and neutralization assay, separately. Methodology FACS & Single-Cell Analysis: B cells from severe COVID-19 patients are sorted based on IFN-α2 binding capacities. Gene Expression Profiling: scRNAseq and differential gene expression (DGE) analysis define autoreactive B cell subsets. AI-Based AutoAb Sequence Analysis: Deep learning models classify autoreactive BCR sequences and IGHV/IGKV/IGLV usage. AlphaFold Modeling & Docking: Predicts high-affinity interactions between autoAbs and IFN-α2. Expected Outcomes A comprehensive AI-driven database of anti-cytokine autoAbs. Identification of high-risk individuals based on immunological markers and their B. cell receptor repertoire. Structural insights into IFN-α2 autoAb interactions for therapeutic targeting. Conclusion By integrating AI, AlphaFold, and scRNAseq, this project aims to systematically characterize anti-cytokine autoAbs, particularly those implicated in severe COVID-19. Our findings will enhance diagnostics, risk stratification, and potential therapeutic interventions for autoimmune and infectious diseases.