I am thrilled to apply for the Hong Kong Laureate Forum, a premier platform uniting emerging scientists with Shaw Laureates to advance global scientific dialogue. As a pharmacy researcher at University of Hong Kong, I am driven by questions like the design of next generation of multicomponent crystals and I believe the Forum’s interdisciplinary ethos will catalyze transformative insights. My work on solid state pharmaceutics honed my skills in computational chemistry. Beyond research, I actively engage in outreach and conferences, reflecting my commitment to science communication—a pillar of the Forum’s mission. The Shaw Prize’s recognition of paradigm-shifting work in Life Science and Medicine inspires me; I am eager to learn from Laureates and discuss challenges. Hong Kong’s role as a science and culture crossroads excites me. The Forum’s workshops and networking opportunities will help me forge collaborations while broadening my perspective on global research trends. I hope to contribute my expertise in solid state pharmaceutics to discussions and take home ideas to address the current challenges. This Forum represents the exact synergy of excellence and mentorship I seek to propel my career. I would be honored to join this cohort of future scientific leaders and appreciate your consideration.

Multicomponent crystals (MCCs), including cocrystals, salts, and hydrates, enhance pharmaceutical properties like solubility and bioavailability. However, early-stage screening and prediction remain challenging. While computational and deep learning (DL) methods have improved cocrystal prediction, broader MCC classification is lacking. To address this, MCC-GCN, a graph convolutional network model, was developed to classify MCCs into cocrystals, salts, hydrates, or failures.

A dataset of 34,618 samples from the Cambridge Structural Database (CSD) was augmented and used for training. The model achieved 80.12% balanced accuracy in cross-validation, outperforming benchmarks. Transfer learning improved predictions for structurally similar APIs like minoxidil, kopexil, and kopyrrol. Experimental validation involved synthesizing MCCs of kopexil and kopyrrol with 32 coformers, yielding two cocrystals, five salts, and 17 hydrates for kopexil, and five cocrystals, nine salts, and eight hydrates for kopyrrol. MCC-GCN correctly predicted 50% of experimental cases.

Molecular electrostatic potential (MEP) analysis revealed that intermolecular interactions drive MCC formation, while hydrogen bond donor/acceptor differences and polar surface area influence crystal type. The study demonstrates MCC-GCN’s practical utility in accelerating MCC development and provides insights into crystal engineering mechanisms. Overall, the model offers a robust tool for pharmaceutical crystal design, supported by experimental validation.