COFAP: A Universal Framework for COFs Adsorption Prediction through Designed Multi-Modal Extraction and Cross-Modal Synergy

2025-11-03

AI Summary

Conventional machine learning methods for high-throughput screening of covalent organic frameworks (COFs) suffer from gas-specific feature dependency, high computational cost, and poor generalizability. Method: This work proposes a universal adsorption performance prediction framework based on deep learning, jointly extracting multimodal structural representations—including atomic graphs, topological descriptors, and chemical sequences—and integrating them via a cross-modal attention mechanism. It eliminates reliance on empirical parameters (e.g., Henry’s coefficients or isosteric heats of adsorption) by adopting an end-to-end architecture and introduces a tunable-weight ranking scheme enabling flexible optimization under physical constraints (e.g., pore size, specific surface area). Contribution/Results: Evaluated on the hypoCOFs dataset, the framework achieves state-of-the-art prediction accuracy and improves inference efficiency by over one order of magnitude, significantly advancing inverse design and large-scale screening of COFs for gas separation applications.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search
• Computer Vision: Learning & Optimization for CV

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Covalent organic frameworks (COFs) are promising adsorbents for gas adsorption and separation, while identifying the optimal structures among their vast design space requires efficient high-throughput screening. Conventional machine-learning predictors rely heavily on specific gas-related features. However, these features are time-consuming and limit scalability, leading to inefficiency and labor-intensive processes. Herein, a universal COFs adsorption prediction framework (COFAP) is proposed, which can extract multi-modal structural and chemical features through deep learning, and fuse these complementary features via cross-modal attention mechanism. Without Henry coefficients or adsorption heat, COFAP sets a new SOTA by outperforming previous approaches on hypoCOFs dataset. Based on COFAP, we also found that high-performing COFs for separation concentrate within a narrow range of pore size and surface area. A weight-adjustable prioritization scheme is also developed to enable flexible, application-specific ranking of candidate COFs for researchers. Superior efficiency and accuracy render COFAP directly deployable in crystalline porous materials.

Problem

• Predicting gas adsorption in COFs without requiring time-consuming experimental features
• Overcoming limitations of conventional ML methods that rely on specific gas properties
• Enabling efficient high-throughput screening of optimal COF structures from vast design space

Innovation

• Extracts multi-modal structural and chemical features
• Fuses complementary features via cross-modal attention
• Enables flexible ranking with weight-adjustable prioritization scheme