In a groundbreaking development for vaccine design and immunotherapy, researchers from the Pritzker School of Molecular Engineering (PME) at The University of Chicago have harnessed the capabilities of machine learning to identify novel immune pathway-enhancing molecules. The study, recently published in the journal Chemical Science, showcases the potential of artificial intelligence (AI) in revolutionizing the search for immunomodulators, crucial components for developing more effective vaccines and robust cancer immunotherapies.
A vast chemical space explored
The challenge in identifying the right molecules to elicit the desired immune response has been formidable, given the staggering estimate of 10^60 drug-like small molecules—far surpassing the number of stars visible in the universe. To navigate this expansive chemical space, the research team, led by Prof. Aaron Esser-Kahn, employed machine learning techniques, a method not previously applied in this manner for immunomodulator discovery.
AI-guided screening process
The team initiated a high-throughput screening of 40,000 molecule combinations to evaluate their impact on innate immune pathways, specifically targeting the NF-κB and IRF pathways crucial for inflammation and antiviral responses. Subsequently, the researchers combined the results with a library of nearly 140,000 commercially available small molecules to guide an iterative computational and experimental process.
Active learning unveils hidden gems
Utilizing active learning, a machine learning technique that efficiently navigates experimental screening through molecular space, graduate student Yifeng (Oliver) Tang led the charge. The process was iterative, with the model suggesting potential candidates or unexplored areas, prompting the team to conduct high-throughput analyses and feed the data back into the active learning algorithm. Astonishingly, after just four cycles, sampling a mere 2% of the library, the team identified previously undiscovered high-performing small molecules.
Record-breaking results
The AI-guided discovery revealed small molecules with record-level performance, surpassing human intuition. These top-performing candidates demonstrated a remarkable 110% improvement in NF-κB activity, an 83% elevation in IRF activity, and a striking 128% suppression of NF-κB activity. One standout molecule exhibited a three-fold enhancement of IFN-β production when delivered with a STING agonist, showcasing promise for cancer treatment.
Generalists and their versatility
The research also uncovered “generalists”—immunomodulators capable of modifying pathways when co-delivered with agonists, chemicals that activate cellular receptors. These versatile small molecules could potentially play a multifaceted role in various vaccines, making them easier to bring to market. Prof. Andrew Ferguson emphasized the excitement surrounding the prospect of a single molecule contributing to a broad spectrum of vaccines.
Unraveling molecular secrets
To gain insights into the characteristics of the identified molecules, the team conducted a thorough analysis of common chemical features that promoted desirable behaviors. This knowledge enables a targeted focus on molecules with specific characteristics or the rational engineering of new molecules with identified chemical groups.
The researchers intend to continue this innovative process, aiming to identify molecules with more specific immune activity and exploring combinations that offer better control of the immune response. Prof. Esser-Kahn expressed the ultimate goal—to find molecules capable of treating diseases.
A paradigm shift in vaccine design
The use of machine learning to guide the discovery of immunomodulators marks a paradigm shift in vaccine design and immunotherapy. The success of this AI-driven approach not only accelerates the identification of potent molecules but also opens avenues for collaboration within the scientific community. As the team looks forward to expanding their search for molecules, they encourage sharing datasets to enhance the efficiency and impact of this transformative research.
Source: https://www.cryptopolitan.com/unleashing-the-power-of-machine-learning/