AI for RNA Biology

The need for Singapore to prioritize understanding and innovation over manufacturing in order to maintain a competitive edge, particularly in knowledge-intensive sectors like pharmaceuticals, was highlighted. The importance of enhancing collaboration among major academic and scientific institutions to maximize impact and reduce internal competition was discussed. Additionally, it was noted that there is a need to reevaluate how scientific progress is measured, and a crucial need for metrics that capture long-term impact. 

AI-based models for RNA research, including RiNalMo, DRAGON, LEAP-R3 and RNA-Ligand complex models, were presented during the workshop. Discussions advocated for leveraging existing datasets like SG100K for future research endeavors.

The need for explainable and interpretable models in biological contexts was highlighted. Multimodal AI integration, combining single-cell RNA profiling with patient metadata and other data types, was mentioned as having high potential to achieve a breakthrough in biomarker discovery. Singapore's role as an early technology adopter was recognized, although concerns were raised about potentially premature adoption without addressing clinical needs and the sustainability in training large models. 

During the panel discussions, the challenges of AI acceptance and adoption by clinicians and regulators was discussed. The importance of benchmarking AI models against established standards and improving patient outcomes was highlighted as a minimum requirement to ensure acceptance and adoption. The importance of developing better means of assessing model’s accuracy, generalizability, and their positive impact on patient experience was also discussed, along with future strategies for integrating AI into healthcare systems and the potential need for changed human behavior before AI can properly be integrated into these systems.