MILP-based copolymer design via mixing vector representation
Researchers extend inverse molecular design to copolymers using convex combinations of monomer descriptors, enabling exact optimization without sequence information.
Source: arxiv/cs.LG · Jianshen Zhu, Raveena Rai, Taiyo Sohkawa, Naveed Ahmed Azam, Kazuya Haraguchi, Liang Zhao, Tatsuya Akutsu · open original ↗
A mixing vector model enables exact copolymer inverse design via MILP by representing polymer properties as weighted combinations of monomer features.
- — Extends mol-infer framework to copolymers using convex combinations of monomer descriptors weighted by composition ratios.
- — Avoids explicit sequence-class information, making the approach naturally compatible with MILP solvers.
- — Achieves R² > 0.7 on nine of ten physicochemical property datasets; six exceed R² > 0.9.
- — Formulates multi-monomer inverse-design problems that remain computationally tractable for three-monomer systems.
- — Validates inferred candidates by re-evaluating predictions against learned model outputs.
- — Provides first tractable step toward exact copolymer design under two-layered model framework.
Astrobobo tool mapping
- Knowledge Capture Document the mixing vector model definition and the MILP formulation for your specific copolymer system. Record monomer descriptors and property datasets as reference.
- Focus Brief Summarize the three-monomer inverse-design workflow: define property targets, set composition constraints, solve MILP, validate candidates. Use this as a checklist for your next design iteration.
- Reading Queue Queue the original mol-infer paper (referenced here) and recent MILP-based molecular design work to understand the two-layered model and solver landscape.
Frequently asked
- The mixing vector model represents a copolymer's properties as a weighted average of its constituent monomer descriptors, where weights are the mole fractions of each monomer. This convex combination avoids the need to encode sequence information and allows the problem to be solved exactly using mixed integer linear programming (MILP).
cite ▸
APA
Jianshen Zhu, Raveena Rai, Taiyo Sohkawa, Naveed Ahmed Azam, Kazuya Haraguchi, Liang Zhao, Tatsuya Akutsu. (2026, May 29). MILP-based copolymer design via mixing vector representation. Astrobobo Content Engine (rewrite of arxiv/cs.LG). https://astrobobo-content-engine.vercel.app/article/milp-based-copolymer-design-via-mixing-vector-representation-2312c1
MLA
Jianshen Zhu, Raveena Rai, Taiyo Sohkawa, Naveed Ahmed Azam, Kazuya Haraguchi, Liang Zhao, Tatsuya Akutsu. "MILP-based copolymer design via mixing vector representation." Astrobobo Content Engine, 29 May 2026, https://astrobobo-content-engine.vercel.app/article/milp-based-copolymer-design-via-mixing-vector-representation-2312c1. Based on "arxiv/cs.LG", https://arxiv.org/abs/2605.29329.
BibTeX
@misc{astrobobo_milp-based-copolymer-design-via-mixing-vector-representation-2312c1_2026,
author = {Jianshen Zhu, Raveena Rai, Taiyo Sohkawa, Naveed Ahmed Azam, Kazuya Haraguchi, Liang Zhao, Tatsuya Akutsu},
title = {MILP-based copolymer design via mixing vector representation},
year = {2026},
url = {https://astrobobo-content-engine.vercel.app/article/milp-based-copolymer-design-via-mixing-vector-representation-2312c1},
note = {Astrobobo rewrite of arxiv/cs.LG, https://arxiv.org/abs/2605.29329},
}