Autonomous Multi-Agent AI for High-Throughput Polymer Informatics: From Property Prediction to Generative Design Across Synthetic and Bio-Polymers

TL;DR

This work introduces a unified multi-agent AI ecosystem for polymer informatics that orchestrates LLM-guided reasoning, graph-based predictors, physics-informed models, and generative design within a Polymer Research Lifecycle. It demonstrates near‑state‑of‑the‑art predictive accuracy on polymer properties ($R^2$ values around 0.8–0.9 for Tg, density, and mechanical properties) with linear scalability to at least 10,000 polymers and low per‑polymer cost. The framework extends to biopolymers with a multimodal protein analysis pipeline and to autonomous polystyrene design with metacognitive self-assessment, highlighting robust cross‑domain applicability. Ablation studies show specialized components (PolyGNN, PINN, validation, knowledge graphs) are essential to achieving high accuracy, robustness, and efficient scaling, while cross-modal validation remains a key challenge. Overall, the approach promises fast, low-cost, high-throughput polymer discovery and design with uncertainty estimates, enabling more autonomous and scalable materials development while underscoring the need for stronger cross‑modal consistency checks and domain grounding.

Abstract

We present an integrated multiagent AI ecosystem for polymer discovery that unifies high-throughput materials workflows, artificial intelligence, and computational modeling within a single Polymer Research Lifecycle (PRL) pipeline. The system orchestrates specialized agents powered by state-of-the-art large language models (DeepSeek-V2 and DeepSeek-Coder) to retrieve and reason over scientific resources, invoke external tools, execute domain-specific code, and perform metacognitive self-assessment for robust end-to-end task execution. We demonstrate three practical capabilities: a high-fidelity polymer property prediction and generative design pipeline, a fully automated multimodal workflow for biopolymer structure characterization, and a metacognitive agent framework that can monitor performance and improve execution strategies over time. On a held-out test set of 1,251 polymers, our PolyGNN agent achieves strong predictive accuracy, reaching R2 = 0.89 for glass-transition temperature (Tg ), R2 = 0.82 for tensile strength, R2 = 0.75 for elongation, and R2 = 0.91 for density. The framework also provides uncertainty estimates via multiagent consensus and scales with linear complexity to at least 10,000 polymers, enabling high-throughput screening at low computational cost. For a representative workload, the system completes inference in 16.3 s using about 2 GB of memory and 0.1 GPU hours, at an estimated cost of about $0.08. On a dedicated Tg benchmark, our approach attains R2 = 0.78, outperforming strong baselines including single-LLM prediction (R2 = 0.67), group-contribution methods (R2 = 0.71), and ChemCrow (R2 = 0.66). We further demonstrate metacognitive control in a polystyrene case study, where the system not only produces domain-level scientific outputs but continually monitors and optimizes its own behavior through tactical, strategic, and meta-strategic self-assessment.

Figures (22)

• Figure 1: Overall architecture of the multi-agent system.
• Figure 2: Sequence coverage plot showing depth of aligned sequences across positions and corresponding sequence identity to the query.
• Figure 3: Ensemble of AlphaFold models 2–5 with different values of pLDDT, pTM, and RMSD_tol.
• Figure 4: (Top) Interactive 3D structure of the predicted protein, for which expert analysis confirms a single-domain architecture. (Bottom) Corresponding Predicted Aligned Error (PAE) plot. The vision agent’s interpretation of these figures was factually incorrect, illustrating a failure in cross-modal verification.
• Figure 5: Property trade-off analysis for sustainable polymer candidates. The parallel-coordinates plot shows how different design choices simultaneously impact multiple target properties, supporting transparent multi-objective decision-making.