UCLA researchers have combined language models with laboratory testing to explore potential ways to overcome resistance to cancer treatment
An AI platform has helped researchers identify a drug combination that could warrant further investigation in prostate cancer, offering a potential way to narrow the search for treatments that work together against resistant disease.
Developed at the University of California, Los Angeles (UCLA), the platform uses large language models to analyse established knowledge about drug actions and the biological pathways cancer cells use to survive treatment.
Called MAESTRO, the system ranks potential combinations and generates biological explanations for researchers to investigate. In a study published in Advanced Science, the team used it to select compounds for testing alongside enzalutamide, a medicine used to treat advanced prostate cancer.
Laboratory experiments identified berberine, a naturally occurring plant compound, as a promising partner. The findings are from cancer cell models and have yet to establish whether the combination would be safe or effective in patients.
The challenge of resistance
Many prostate cancers depend on androgen receptor signalling to grow. Enzalutamide blocks this signalling, but cancer cells can adapt by activating alternative pathways that support their survival. Identifying drugs that could disrupt those escape routes requires researchers to assess numerous compounds and interconnected biological processes.
“Developing drug combinations currently relies on either brute-force screening or researchers hypothesizing potential candidates based on existing literature,” says Neil Lin, Associate Professor of Mechanical and Aerospace Engineering and Bioengineering at UCLA and the study’s senior author.
MAESTRO draws together information on prostate cancer biology, resistance mechanisms and known drug actions to prioritise combinations with a biological rationale for testing.
From prediction to laboratory testing
The researchers used the platform to evaluate 3,592 natural products, selecting 150 high-confidence candidates for experiments in 3D prostate cancer spheroids. These small clusters of cancer cells allow researchers to examine aspects of tumour biology beyond those captured by cells grown in flat layers. The study used oxygen-supplemented spheroids to reduce the influence of oxygen deprivation on experimental results.
Rather than relying on a single prediction, the team evaluated each candidate repeatedly, considering both its average ranking and the consistency of the results. Of the 150 compounds tested, 23 showed evidence of synergy with enzalutamide, producing a greater effect than expected from the individual treatments.
Further experiments identified consistent synergy between berberine and enzalutamide across four prostate cancer models, including both treatment-resistant and treatment-sensitive models.
A testable biological explanation
Researchers also examined whether the combination’s cellular effects were consistent with the explanation generated by MAESTRO.
The pairing increased activity associated with AMPK, which helps regulate cellular energy and stress, while reducing signalling associated with mTORC1, a regulator of cell growth and survival. It was also linked to increased cellular stress and reduced DNA replication and cell-cycle progression.
These findings support further investigation of the proposed biological explanation, rather than establishing the complete mechanism behind the combination’s effects.
The team plans to evaluate the pairing in patient-derived tumour models and animal studies. Researchers also intend to incorporate additional biological data into MAESTRO and explore its application to other drug combinations, resistance mechanisms and cancer types.
The platform’s immediate contribution is to help scientists decide which experiments to undertake, connecting existing biomedical knowledge with a more focused programme of laboratory testing.
Have some thoughts on the topic? Share them with the community here