Revealing the Molecular Architecture of PWS Through Large Language Models for Targeted Drug Repurposing

Funding Summary

Dr. Singh will apply artificial intelligence to PWS datasets to glean new information about pathways disrupted in PWS and possible targets for therapy.

Dr. Theresa Strong, Director of Research Programs, shares details on this project in this short video clip.

Lay Abstract

Summary of Project Aims and Research Plan
We are embarking on a transformative project to tackle Prader-Willi Syndrome (PWS) through a cutting-edge combination of artificial intelligence (AI) and advanced genomics. Our team aims to create a comprehensive map of how the deletion of certain genes in PWS affects other genes and pathways within cells. By integrating a wealth of data from both general health and PWS-specific sources, we will use powerful computer models to predict critical interactions and identify key points where therapeutic interventions could be effective. Additionally, we plan to scan thousands of existing drugs, as well as millions of potential drugs, to find those that could potentially be applied to treat PWS, significantly speeding up the discovery of treatments.
Importance of the Research
Our research is crucial because it uses innovative methods to address a gap in current treatment options for PWS. The integration of AI with detailed genetic and protein data promises to revolutionize our understanding of PWS at a systems level, identifying new drug targets and accelerating the repurposing of existing drugs. This approach not only enhances our biological understanding but also opens up faster, more cost-effective paths to treatment compared to traditional drug development. It will also enable PWS researchers to efficiently test their hypotheses by identifying the commercially-available molecules most likely to target their protein of interest.
Expected Impact and Next Steps
In the short term (2-5 years), we expect to deliver an integrated map of the PWS subsystem that will incorporate gene expression, protein interactions, and non-coding RNAs. We anticipate identifying multiple potential drug targets and existing drugs that could be redirected to treat PWS. These findings could lead directly to clinical trials and potentially to new, effective treatments soon afterward.
Advancing Therapeutic Development and Clinical Care
This project directly advances therapeutic development by pinpointing precise intervention points within the disease’s protein networks and providing a rapid screening method to identify viable drugs for treatment. By starting with drugs already approved for other conditions, we can quickly move successful candidates into clinical trials, offering new hope for effective treatments; we will then scan much more broadly for other treatment possibilities.

Research Outcomes: Public Summary

Prader-Willi syndrome (PWS) is driven by a network of genes that extends well beyond the chromosome 15q11-q13 region affected in the disorder. A central goal of this project was to map that broader network computationally and turn it into concrete, testable leads for treatment. The work produced two main advances for the PWS community.
A genome-wide map of PWS-relevant genes — and the pathways they act through. Starting from known PWS genes, our prioritization framework (GLOWgenes++) identified roughly 80 additional genome-wide candidate genes with strong functional connections to PWS biology. When we grouped these genes by function, they concentrated in exactly the pathways that underlie the hallmark features of PWS: insulin and IGF signaling (INSR, IRS2, and the IGF-binding proteins IGFBP1–4 and IGFALS), glucose transport and metabolism (SLC2A4, NAMPT), adipokine and metabolic regulation (ADIPOR2, C1QTNF9), neuropeptide and appetite regulation (AGRP, AVP), catecholamine and neurotransmitter synthesis (TH, TYR) with adrenergic signaling (ADRA2B), thyroid hormone metabolism (DIO3), and growth-factor and neuronal support (NGF). These broader target pathways align closely with the metabolic, hormonal, and hyperphagia phenotypes that define PWS, giving researchers a prioritized, mechanism-anchored set of genes and pathways to investigate rather than a single locus.
Candidate drugs nominated for experimental testing. Building on this gene map, we developed a computational drug-screening pipeline that searches for compounds predicted to reverse the PWS molecular signature and to bind PWS-relevant targets. The pipeline nominated specific candidate drug–target pairs for follow-up including: Luminespib (targeting the serotonin receptor 5-HT1A) and Norepinephrine (targeting the dopamine receptor DRD3). alongside a broader set of candidate compounds drawn from approved-drug and large chemical libraries. These are computational predictions intended as starting points for laboratory validation in partnership with experimental PWS researchers, not treatments.
Open resources for the community. To make these findings reusable, we are releasing the underlying tools so other PWS researchers can apply and build on them: the GLOWgenes++ gene-prioritization code [https://github.com/rohitsinghlab/rdgenefinder], and the reverse-signature drug-search web app [available from the Github repository]. Together, these results move PWS target discovery toward a systems-level framework: a prioritized map of disease-relevant genes and pathways, a set of candidate drugs ready for validation, and open tools the community can use.