Track genomics, neuroscience and molecular biology.
Search papers, labs, and topics across Lattice.
This paper introduces three testable criteria—generality, transferability, and generalization—to rigorously evaluate the "foundational" nature of current molecular models, moving beyond vague descriptors. Application of these criteria to state-of-the-art models reveals significant progress in areas like biomolecular structure prediction and interatomic potentials, but no single model fully meets all criteria. The findings underscore that data quality, structure, and the integration of prior knowledge are more critical for advancement than model scale alone, especially in domains with consistent property definitions and abundant, low-noise data.
Current molecular foundation models excel in niche areas but fall short of true generality, transferability, and out-of-distribution generalization, highlighting data and prior knowledge integration as key to future progress.
Large-scale models have permeated the molecular sciences, yet what makes a model'foundational'in this domain remains poorly defined. This paper proposes three testable criteria for assessing the foundational nature of molecular models: (i) generality across molecular entities, properties, and tasks; (ii) transferability to new applications with no or minimal task-specific retraining; and (iii) generalization beyond the training distribution. Applying these criteria to the state of the art reveals promising progress, particularly visible in biomolecular structure prediction and machine-learned interatomic potentials, although none of the approaches examined fully satisfies all three. Success is concentrated in domains where target properties are consistently defined and training data are abundant, with low label noise relative to physically meaningful variation. More broadly, progress in the molecular sciences appears to depend less on model scale alone than on the quality and structure of available data, as well as the incorporation of prior knowledge into models, prediction tasks, or downstream applications. This work shifts the notion of a molecular foundation model from a descriptive label to a testable hypothesis, offering a framework for assessing current models and guiding future developments.