← Back to all news
arXiv AI By Wenda Wang, Yang Zhang, Zhewei Wei, Wenbing Huang

AbFlow : End-to-end Paratope-Centric Antibody Design by Interaction Enhanced Flow Matching

Read the original on arXiv AI →

AbFlow is a one‑step flow‑matching framework that designs full‑atom antibodies end‑to‑end, focusing on the paratope region. It uses an equivariant Surface Multi‑channel Encoder to incorporate surface‑level antigen interaction data, refining especially the CDR‑H3 region. Experiments demonstrate that AbFlow generates superior antigen‑antibody complexes with improved binding affinity, particularly at the contact interface.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv Machine Learning
Sep 4

DuaDeep-SeqAffinity: Dual-Branch Deep Learning for Tri-Stream Sequence-Based Antibody--Antigen Affinity Prediction

DuaDeep-SeqAffinity is a sequence-only deep learning framework that predicts antibody–antigen binding affinity directly from primary amino acid sequences, eliminating the need for resolved 3D structures. The model processes the antigen and the antibody heavy and light chains as three independent streams, each embedded with a frozen ESM‑2 protein language model and passed through parallel Transformer and CNN branches before late fusion. On a sequence‑disjoint split of the AbRank benchmark, it achieves a Pearson correlation of 0.683, an R² of 0.460, and a pairwise ranking AUC of 0.895, outperforming single‑branch ablations and showing attention to CDR loops and epitope residues.

By Aicha Boutorh, Soumia Bouyahiaoui, Manel Kara Laouar, Sara Belhadj, Nour El Yakine Guendouz, Asma Boutorh
llmsefficiencybenchmarks
More like this →