We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Quantitative Biology > Quantitative Methods

Title: Molformer: Motif-based Transformer on 3D Heterogeneous Molecular Graphs

Abstract: Procuring expressive molecular representations underpins AI-driven molecule design and scientific discovery. The research to date mainly focuses on atom-level homogeneous molecular graphs, ignoring the rich information in subgraphs or motifs. However, it has been widely accepted that substructures play a dominant role in the identification and determination of molecular properties. To address such issues, we formulate heterogeneous molecular graphs (HMGs), and introduce Molformer to exploit both molecular motifs and 3D geometry. Specifically, we extract functional groups as motifs for small molecules and resort to the reinforcement learning to adaptively select quaternary amino acids as motifs for proteins. Then HMGs are constructed with both atom-level and motif-level nodes. To better accommodate those HMGs, we introduce a variant of Transformer named Molformer, which adopts a heterogeneous self-attention layer to distinguish the interactions between multi-level nodes. Besides, it is also coupled with a multi-scale mechanism to capture local fine-grained patterns with increasing contextual scales. An attentive farthest point sampling algorithm is also proposed to obtain the molecular representations. We validate Molformer across a few domains including quantum chemistry, physiology, and biophysics. Experiments show that Molformer outperforms state-of-the-art baselines. Our work provides a promising way to utilize informative motifs from the perspective of multi-level graph construction.
Subjects: Quantitative Methods (q-bio.QM); Computational Engineering, Finance, and Science (cs.CE); Machine Learning (cs.LG)
Cite as: arXiv:2110.01191 [q-bio.QM]
  (or arXiv:2110.01191v5 [q-bio.QM] for this version)

Submission history

From: Fang Wu [view email]
[v1] Mon, 4 Oct 2021 05:11:23 GMT (9942kb,D)
[v2] Fri, 8 Oct 2021 05:59:37 GMT (9943kb,D)
[v3] Thu, 23 Dec 2021 03:17:08 GMT (9942kb,D)
[v4] Sun, 15 May 2022 11:27:25 GMT (1145kb,D)
[v5] Wed, 18 May 2022 08:10:34 GMT (1110kb,D)

Link back to: arXiv, form interface, contact.