Can agents reliably design therapeutic antibodies?
Abstract
Recent work has demonstrated agents driving protein design workflows, either through orchestration of specialized tools or direct manipulation of sequences. Yet the fidelity with which agents follow protein design instructions and the factors contributing to their effectiveness remain understudied. We examine agentic antibody design under two settings: when explicit molecular-editing instructions are provided and when agents receive only previously-scored sequences and are prompted to propose improvements. We call the first task-guided and the second black-box design. For the task-guided setting, we curate ten expert-validated affinity-improvement tasks and show that LLM judges closely reproduce programmatic evaluations of instruction following. We then study two augmentations: access to antibody-relevant tools and to previously scored design history. Tools improve instruction following (IF) and sequence space exploration while reducing optimization performance, whereas design history reduces IF but improves optimization performance at the expense of exploration. Finally, we combine taskguided and black-box design within an in silico optimization loop and show that our agentic optimizer performs comparably to an established Bayesian optimization baseline.
Related
- Therapeutic anti-IgE antibodies generate distinct functional immune complex signatures
- Structure and function of therapeutic antibodies approved by the US FDA in 2025
- A cell death screen identifies macrophage-depleting agents with therapeutic potential
- Evaluation of vaccine-induced anti-lipopolysaccharide antibodies against clinically relevant Klebsiella pneumoniae strains to inform vaccine design
- ADPKD as a Temporally Dynamic Disease: Implications for Biomarkers, Trial Design, and Therapeutic Targeting
BibTeX
@misc{alpman2026agents,
title = {{Can agents reliably design therapeutic antibodies?}},
author = {Asli Alpman and Nick Bhattacharya and Aaron Solomon and Hunter Elliott},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.00558},
eprint = {2611.00558},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.00558},
}
RIS
TY - UNPB
TI - Can agents reliably design therapeutic antibodies?
AU - Asli Alpman
AU - Nick Bhattacharya
AU - Aaron Solomon
AU - Hunter Elliott
PY - 2026
DA - 2026/11/01
PB - arXiv
DO - 10.48550/arxiv.2611.00558
UR - https://arxiv.org/abs/2611.00558
AB - Recent work has demonstrated agents driving protein design workflows, either through orchestration of specialized tools or direct manipulation of sequences. Yet the fidelity with which agents follow protein design instructions and the factors contributing to their effectiveness remain understudied. We examine agentic antibody design under two settings: when explicit molecular-editing instructions are provided and when agents receive only previously-scored sequences and are prompted to propose improvements. We call the first task-guided and the second black-box design. For the task-guided setting, we curate ten expert-validated affinity-improvement tasks and show that LLM judges closely reproduce programmatic evaluations of instruction following. We then study two augmentations: access to antibody-relevant tools and to previously scored design history. Tools improve instruction following (IF) and sequence space exploration while reducing optimization performance, whereas design history reduces IF but improves optimization performance at the expense of exploration. Finally, we combine taskguided and black-box design within an in silico optimization loop and show that our agentic optimizer performs comparably to an established Bayesian optimization baseline.
ER -