Having better control over the responses of immune signaling pathways is vital in prophylactic vaccines to reduce systemic inflammation, thus, ensuring better safety and efficacy. We previously found that immunomodulators co-delivered with certain Pattern Recognition Receptor (PRR) agonists could alter the responses stimulated by agonists alone, enabling us to fine-tune the immune activity as needed. In this work, we attempted to expand the library of immunomodulators with small molecules that can modulate two major effectors of innate immune responses, the nuclear factor kappa-light-chain-enhancer of activated B-cells (NF-κB) pathway and the interferon regulatory factors (IRF) pathway, which are drawn from selected chemical screening libraries. However, the vast size of the chemical space makes it intractable to screen every molecule through all the libraries by cell assays. Hence, we developed a data-driven pipeline integrating machine learning (ML) and high throughput screening (HTS) to search for molecules in the library having the desired modulation on the immune pathways. We encoded molecular structure into continuous latent embeddings using Variational Auto-encoder (VAE), then proposed candidate molecules using Bayesian Optimization (BO). Compared to computer-assisted screening pipelines where only one biochemical property is being optimized, we were looking at a multi-objective problem and seeking to discover molecules that can either enhance or inhibit specific innate immune responses. By doing so, we can efficiently traverse the small molecule libraries and offer massive savings in time, money, and labor by rationally directing the experimental screen. We identified a set of molecules that can either up-regulate or down-regulate the NF-κB or IRF activity by more than 15 times. We further validated a sub-sample of the molecules discovered by the framework and found that they can also have a large-scale modulation effect on the cytokine release profile. These molecular candidates can be beneficial to the development of vaccine adjuvants by ensuring better control over the level of inflammation and to immunotherapy by enhancing the immune stimulation to overcome the tumor micro-environmental suppression. The data-driven framework can also apply to many early-stage drug-discovery pipelines to provide the ability to search a much vaster chemical space than traditional high throughput screening could handle.

Data-driven discovery of innate immunomodulators via a closed-loop system combining machine learning and high throughput screening

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.