Multi-Hierarchical Fine-Grained Feature Mapping Driven by Feature Contribution for Molecular Odor Prediction

TL;DR

This work tackles molecular odor prediction by addressing expressive feature limitations and severe descriptor imbalance. It introduces HMFNet, a multi-hierarchical framework combining Local Multi-Hierarchy Feature Extraction (LMFE) with Harmonic Modulated Feature Mapping (HMFM) and Global Multi-Hierarchy Feature Extraction (GMFE), all guided by a Chemically-Informed Loss (CIL). LMFE captures fine-grained atomic/bond information with feature-importance and frequency-modulation pathways, while GMFE leverages global fingerprints and SMILES-based representations, fused through dual MLPs and a GAT-based encoder. CIL integrates weighted BCE, energy-based constraints, sample-level regulation, and label-correlation terms to improve minority-class predictions and capture descriptor co-occurrences. Empirically, the method yields state-of-the-art F1 and competitive AUROC across baseline models, demonstrating improved molecular structure-to-odor mappings with potential impact on fragrance design and related applications.

Abstract

Molecular odor prediction is the process of using a molecule's structure to predict its smell. While accurate prediction remains challenging, AI models can suggest potential odors. Existing methods, however, often rely on basic descriptors or handcrafted fingerprints, which lack expressive power and hinder effective learning. Furthermore, these methods suffer from severe class imbalance, limiting the training effectiveness of AI models. To address these challenges, we propose a Feature Contribution-driven Hierarchical Multi-Feature Mapping Network (HMFNet). Specifically, we introduce a fine-grained, Local Multi-Hierarchy Feature Extraction module (LMFE) that performs deep feature extraction at the atomic level, capturing detailed features crucial for odor prediction. To enhance the extraction of discriminative atomic features, we integrate a Harmonic Modulated Feature Mapping (HMFM). This module dynamically learns feature importance and frequency modulation, improving the model's capability to capture relevant patterns. Additionally, a Global Multi-Hierarchy Feature Extraction module (GMFE) is designed to learn global features from the molecular graph topology, enabling the model to fully leverage global information and enhance its discriminative power for odor prediction. To further mitigate the issue of class imbalance, we propose a Chemically-Informed Loss (CIL). Experimental results demonstrate that our approach significantly improves performance across various deep learning models, highlighting its potential to advance molecular structure representation and accelerate the development of AI-driven technologies.

Figures (6)

• Figure 1: (a): The concrete structure of fine-grained Local Multi-Hierarchical Feature Extraction; (b): The concrete structure of Global Multi-Hierarchy Feature Extraction; (c): The concrete structure of concrete structure of Harmonic Modulated Feature Mapping; (d): The architecture of the proposed overall framework.
• Figure 2: Co-ocurrence matrix of Top 20 odor descriptors.
• Figure 3: Comparison of F1 scores of histogram of Harmonic Modulated Feature Mapping on mainstream deep learning model.
• Figure 4: Comparison of histogram AUROC of Harmonic Modulated Feature Mapping on mainstream deep learning model.
• Figure 5: Distribution of odor descriptor frequency in dataset.