Newest papers

Fragment-Level Graph Encoding for Interpretable Molecular Property Prediction

Rafael Lopes Almeida, Ina Pöhner · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29589 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Graph neural single networks (GNNs) adalah salah satu alat standar untuk prediksi properti molekuler, tetapi penjelasan tingkat atom dari prediksi mereka sulit bagi ahli kimia medis untuk bertindak. Kami mengusulkan perubahan representasi input daripada model baru: molekul dikodekan sebagai grafik tingkat fragmen dari dekomposisi BRICS, di mana setiap node adalah fragmen sintetis yang bermakna, yang fiturnya adalah jumlah unsur-unsur karakteristik atomnya. Karena hanya perubahan representasi, setiap pesan-passing GNN dan setiap metode atribusi post-hoc menghasilkan penjelasan pada resolusi kelompok fungsional alih-alih atom tunggal. Melalui empat arsitektur, empat tugas prediksi, dan dua penjelasan di bawah split, fragment-level encoding cocok dengan akurasi tes tingkat atom pada graf yang kira-kira lima kali lebih kecil. Integrated Gradient attributions cost

Machine translation (Meta M2M100); the English original follows.

Abstract

Graph neural networks (GNNs) are among the standard tools for molecular property prediction, but the atom-level explanations of their predictions are hard for medicinal chemists to act on. We propose a change of input representation rather than a new model: molecules are encoded as fragment-level graphs from BRICS decomposition, where each node is a synthetically meaningful fragment whose features are the element-wise sum of its atoms' features. Because only the representation changes, any message-passing GNN and any post-hoc attribution method produces explanations at the resolution of functional groups instead of single atoms. Across four architectures, four prediction tasks, and two explainers under scaffold splits, fragment-level encoding matches atom-level test accuracy on roughly five-fold smaller graphs. Integrated-Gradients attributions recover known pharmacophores above chance on two enzyme targets, namely BACE at 1.5 times the chance rate, and PTR1 at 3.2 times. A like-for-like comparison against an aggregated atom-level baseline shows that post-hoc aggregation is not a stable substitute: its BACE hit rate range both exceeds and falls below the fragment model's hit rate depending on two arbitrary aggregation choices, whereas fragment-level training yields a single convention-free attribution at no predictive cost. The code is available at: https://github.com/4D-Lab/frame

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Property (philosophy), Fragment (logic), Encoding (memory), Computer science, Representation (politics), Attribution

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange