How AI Is Transforming Drug Discovery and Protein‑Misfolding Research
Advanced computational tools and structural modeling are redefining how scientists investigate protein misfolding and accelerate therapeutic innovation
Artificial intelligence is becoming the most influential force in modern life sciences research, particularly in drug discovery and protein misfolding studies.
In a panel discussion at the Hong Kong Laureate Forum in Hong Kong on November 6, 2025, two renowned molecular biologists—Peter Walter and Kazutoshi Mori, both 2014 Shaw Laureates in Life Science and Medicine—outlined how AI is reshaping laboratory practice while stressing that biological rigor remains indispensable. The panel discussion was moderated by Danny Chan, Emeritus Professor, School of Biomedical Sciences, The University of Hong Kong.
Walter opened the discussion with a clear assessment of AI’s strengths and limitations.
“Probably 95% of the data out there is crap,” he said. “AI will not be able to distinguish between crappy data and good data, and whatever comes out of it has to be put to the experimental test.”
He stressed that biological insight—not computation—remains the foundation of meaningful discovery, adding that “these stress pathways are only activated in cells that are unhappy,” a feature that makes them powerful therapeutic targets.
Building on this point, Walter explained how AI can support scientific reasoning by scanning vast datasets and surfacing associations that may not be obvious at first glance. He noted that this analytical scale enables researchers to explore new hypotheses more rapidly.
However, he emphasized that no model can replace empirical validation, and every computational insight must be tested through direct experimentation before it can guide drug‑development decisions.
Mori added a practical perspective shaped by clinical demand. He explained that physicians rely heavily on stress‑response markers identified through fundamental research to evaluate disease mechanisms.
While AI tools such as AlphaFold (developed by Google DeepMind) have accelerated progress in protein‑folding studies, he noted that laboratory conditions differ widely across research groups. Because of this variability, he said, AI‑generated predictions must always be confirmed with rigorous, systematic experimentation to ensure accuracy.
As laboratories expand their use of AI‑driven screening and analysis, both speakers emphasized the importance of consistent, precise documentation. They noted that clear records of cell types, timing, and environmental conditions allow AI systems to interpret experimental findings accurately and support more reliable decision‑making in drug‑discovery workflows.
Digital Cell Models
During the panel discussion, Walter described the emerging aspiration to build a digital representation of a cell. He highlighted how modern tomograms “contain information about where every component in the cell is,” but admitted that scientists “just can’t distinguish them yet unless they’re big.”
Despite AI’s ability to process massive datasets, he warned that gaps in biological understanding—such as missing phosphatases or small regulatory events—still prevent researchers from assembling a complete virtual cell.
Mori added that this variability extends far beyond protein research. He observed that even subtle differences in protocol, timing, or cell‑culture environments can complicate computational modeling and lead to conflicting results.
To reduce these inconsistencies, he stressed the importance of documenting experimental conditions in detail so AI systems can interpret findings with the necessary context.
Both scientists agreed that while a fully realized digital cell remains a distant goal, partial models—such as receptor‑signaling maps or transcriptional‑program predictors—will increasingly support early‑stage drug‑discovery decisions.
Cellular Decision Logic
Walter addressed the recent trend of describing cells as “intelligent,” especially as AI becomes more prevalent in biological research.
“A cell is a collection of molecules that all obey the rules of thermodynamics,” he said.
He explained that what appears to be cellular decision‑making—such as survival versus programmed cell death—is driven by deterministic biochemical pathways rather than cognition.
Mori noted that researchers are still working to identify the molecular switch that determines whether a cell continues to survive or initiates programmed death under endoplasmic‑reticulum stress. He said that clarifying this mechanism could eventually help scientists influence stress‑response outcomes in disease settings, improving therapeutic precision.
Their combined message was clear: AI can help map causal relationships, but it must operate on biochemically grounded assumptions rather than metaphors of intelligence.
Misfolding & Therapies
Walter emphasized that protein misfolding remains one of the most persistent challenges in understanding neurodegenerative diseases. He noted that subtle shifts in a cell’s environment can dramatically change how proteins fold, making experimental outcomes difficult to predict. These uncertainties make AI‑supported analysis especially valuable, as it can highlight patterns across large datasets that individual experiments may not reveal.
Walter also explained that stress‑response pathways provide a promising route for targeted therapy, because unhealthy or stressed cells activate these pathways far more than healthy ones. He described how selectively inhibiting specific branches of these pathways can eliminate diseased cells with minimal toxicity.
He also pointed to a small molecule from his laboratory that is now advancing through preclinical and clinical evaluation for its potential to correct cognitive defects across multiple neurological diseases.
Mori added that understanding misfolding pathways and stress responses will help clinicians interpret disease markers more accurately, providing a clearer foundation for diagnosis and treatment planning.
Guidance for Researchers
Walter offered a candid reflection on how young scientists should manage the pressure in the early stages of their careers.
“Think of what the cell would do,” he said. “It will turn on stress pathways and try to bring things back into balance.”
He used this analogy to encourage researchers to step back, regain perspective, and avoid falling into cycles of nonstop activity. He recalled trainees who “kept on working like hamsters running on a wheel,” emphasizing that genuine progress comes from clarity of thought, not constant motion.
Mori provided a complementary perspective and shared more of his own experience.
“Find your own strategy to make progress,” he said.
He explained that his turning point came when he began reading widely and regularly, allowing new ideas to “come down to me” after long periods of groundwork.
He said that exposing himself to new papers each month broadened his thinking and helped him recognize important opportunities more clearly. He encouraged young researchers to train their minds deliberately, noting that “you can’t produce anything from everything.”
Together, the laureates urged emerging scientists to cultivate habits that build resilience, creativity, and long‑term direction. Their combined message underscored the importance of thoughtful planning, open‑minded learning, and maintaining a balanced outlook—qualities that sustain a scientist throughout a demanding career.