Learning Molecular Representations from Cellular Phenotypes with Structure Preservation

Multimodal contrastive learning for integrating molecular representations and cellular phenotypes in drug-target interaction prediction

A two-stage contrastive learning framework integrating drug structures, protein sequences, and Cell Painting morphological profiles into a unified embedding space, which reveals pathway-specific morphological signatures associated with drug targets, providing biologically interpretable insights into drug mechanisms.

Self-Supervised Learning for Molecular Property Prediction: Methods, Multimodal Insights, and Benchmark Comparisons

This review provides a systematic overview of recent advances in SSL-based molecular property prediction and analyzes how multimodal molecular representation learning by integrating sequence, graph, three-dimensional structure, and textual information can improve the quality and expressiveness of molecular representations.

Conditional Neural Optimal Transport for Predicting Cellular Phenotypes from Molecular Structure

High-content microscopy enables systematic profiling of cellular responses to chemical perturbations, but the scale of the chemical space makes exhaustive phenotypic characterization experimentally infeasible. This motivates computational models that can predict image-derived phenotypes without acquiring the corresponding treated cells. We formulate molecule-induced phenotype prediction as an inductive conditional transport problem in image representation space. Given a negative-control phenotype and the structure of a molecule, we aim to predict the phenotype induced by the corresponding molecule. We first evaluate classical optimal transport baselines and show that static couplings do not yield useful predictions on large-scale phenotypic image datasets. We then introduce a molecule-conditioned Neural Optimal Transport (NOT) model with a Monge-Gap regularization training objective that learns to transport negative-control unperturbed phenotypes toward perturbed phenotypes using molecular structure as conditioning information. NOT recovers molecule-specific phenotypic effects while reducing microscopy-associated technical variation, thereby facilitating comparisons across experimental batches. On unseen active molecules, the model outperforms baseline approaches, demonstrating that chemically conditioned transport can generalize beyond the molecules observed during training. We identified the molecular encoder as the main limitation to this generalization, while transport in a compressed representation space improves performance and scalability. These results establish NOT as a promising framework for predicting cellular phenotypes from molecular structure and negative-control phenotypes, while highlighting the development of more informative molecular representations as a key direction for improving out-of-distribution performance.

From Cellular Responses to Pharmacological Domains: Multimodal Zero-Shot Drug Representation Learning

PMRD separates mechanism-consistent factors from modality-specific information and constructs a consensus response domain across three modalities and combines complementary representations through reliability-aware multiview retrieval and supports PMRD as an effective framework for mechanism-aware multimodal drug representation learning.

CBInformax: bioactivity-aware self-supervised molecular representation learning for molecular property and drug-drug interaction prediction.

MG-CMIF: A Multi-Granularity Enhanced Cross-Modal Information Fusion Framework for Molecular Property Prediction

Molecular property prediction provides an important computational basis for compound screening and drug development by estimating physicochemical characteristics and biological activities from molecular structures. Although deep learning has improved molecular modeling, existing methods often describe molecules through a limited structural view or combine multiple views without sufficiently exploiting their complementary relationships. In addition, graph-based approaches commonly concentrate on local atomic connectivity, making it difficult to represent chemically meaningful structural units that may strongly influence molecular properties. This paper develops MG-CMIF, a multi-granularity cross-modal framework for molecular property prediction. The proposed model describes each molecule from symbolic, topological, and spatial perspectives and learns an integrated representation through three key designs. First, hierarchical graph modeling combines detailed atomic interactions with substructure-level chemical patterns to enrich topology-oriented features. Second, interaction across molecular views enables information relevant to property prediction to be exchanged selectively rather than merged through shallow operations. Third, alignment-oriented training objectives encourage representations derived from the same molecule to preserve compatible chemical semantics during fusion. Experiments on multiple public benchmark datasets show that MG-CMIF achieves better prediction results than competitive methods in both classification and regression settings. Further ablation analyses confirm that hierarchical structural modeling and cross-view integration both contribute to the effectiveness of the proposed framework.