# Foundation and Multimodal Models for Drug Discovery in Molecular Informatics: Principles, Evaluation, and Practical Guidance

## TL;DR

This paper reviews how foundation and multimodal models are used in drug discovery, offering guidance on their application and evaluation.

## Contribution

The paper introduces practical guidance for using foundation models in drug discovery, emphasizing multimodal integration and rigorous evaluation.

## Key findings

- Foundation models can learn transferable representations from molecular data, aiding drug discovery tasks.
- Multimodal models integrate molecular structures, biological context, and text to improve property prediction and design.
- Rigorous evaluation methods are essential to address distribution shifts and uncertainty in drug discovery.

## Abstract

Foundation and multimodal models are rapidly becoming a core methodology in molecular informatics, particularly for drug discovery, by leveraging large‐scale pretraining across sequences, graphs, 3D structures, and text. This mini‐review provides practical guidance on when these models help, how to choose representations and data, and how to design pretraining and adaptation pipelines for real‐world use. We clarify what qualifies as a foundation model in chemistry; compare chemical language models, graph‐based architectures, and 3D equivariant networks; review multimodal strategies that connect molecules with proteins, pockets, and natural language; and summarize diffusion‐based generative modeling. We also emphasize rigorous evaluation, discussing realistic splitting protocols, distribution shift, activity cliffs, uncertainty calibration, and conformal prediction in the context of widely used benchmarks.

The graphical abstract summarizes how foundation models pretrained on large molecular data provide transferable representations. Through multimodal learning, these representations integrate molecular structure, biological context, and text‐derived knowledge, enabling downstream drug discovery tasks such as property prediction, target interaction modeling, and molecular design.© 2026 WILEY‐VCH GmbH