# Multimodal Protein Language Models for Enzyme Kinetic Parameters: From Substrate Recognition to Conformational Adaptation

arXiv: https://arxiv.org/abs/2603.12845 · HTML: https://mlsanity.com/paper/2603.12845/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 22 of its 76 references that are in the map.
- First posted on arXiv 13 Mar 2026.

## Builds on (most cited first, 22 of 22)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Analyzing Learned Molecular Representations for Property Prediction](https://mlsanity.com/paper/1904.01561/) (2019)
- [Mind the Class Weight Bias: Weighted Maximum Mean Discrepancy for Unsupervised Domain Adaptation](https://mlsanity.com/paper/1705.00609/) (2017)
- [SMILES Transformer: Pre-trained Molecular Fingerprint for Low Data Drug Discovery](https://mlsanity.com/paper/1911.04738/) (2019)
- [PAC-Bayesian Theory Meets Bayesian Inference](https://mlsanity.com/paper/1605.08636/) (2016)
- [Ankh: Optimized Protein Language Model Unlocks General-Purpose Modelling](https://mlsanity.com/paper/2301.06568/) (2023)
- [Kernel Methods on Riemannian Manifolds with Gaussian RBF Kernels](https://mlsanity.com/paper/1412.0265/) (2014)
- [PoET: A generative model of protein families as sequences-of-sequences](https://mlsanity.com/paper/2306.06156/) (2023)
- [Frequency Decoupling for Motion Magnification via Multi-Level Isomorphic Architecture](https://mlsanity.com/paper/2403.07347/) (2024)
- [EulerMormer: Robust Eulerian Motion Magnification via Dynamic Filtering within Transformer](https://mlsanity.com/paper/2312.04152/) (2023)
- [Temporal-Frequency State Space Duality: An Efficient Paradigm for Speech Emotion Recognition](https://mlsanity.com/paper/2412.16904/) (2024)
- [Exploiting Ensemble Learning for Cross-View Isolated Sign Language Recognition](https://mlsanity.com/paper/2502.02196/) (2025)
- [Clustering for Protein Representation Learning](https://mlsanity.com/paper/2404.00254/) (2024)
- [ProtCLIP: Function-Informed Protein Multi-Modal Learning](https://mlsanity.com/paper/2412.20014/) (2024)
- [Multimodal machine learning with large language embedding model for polymer property prediction](https://mlsanity.com/paper/2503.22962/) (2025)
- [Task-Generalized Adaptive Cross-Domain Learning for Multimodal Image Fusion](https://mlsanity.com/paper/2508.15505/) (2025)
- [Pre-train, Align, and Disentangle: Empowering Sequential Recommendation with Large Language Models](https://mlsanity.com/paper/2412.04107/) (2024)
- [Multimodal Pre-Training Model for Sequence-based Prediction of Protein-Protein Interaction](https://mlsanity.com/paper/2112.04814/) (2021)
- [Ankh3: Multi-Task Pretraining with Sequence Denoising and Completion Enhances Protein Representations](https://mlsanity.com/paper/2505.20052/) (2025)
- [Debiased distributed learning for sparse partial linear models in high dimensions](https://mlsanity.com/paper/1708.05487/) (2017)
- [Multimodal Regression for Enzyme Turnover Rates Prediction](https://mlsanity.com/paper/2509.11782/) (2025)

## Built on by (most cited first, 2 of 2)

- [3D Smoke Scene Reconstruction Guided by Vision Priors from Multimodal Large Language Models](https://mlsanity.com/paper/2604.05687/) (2026)
- [Beyond Shortcuts: Mitigating Visual Illusions in Frozen VLMs via Qualitative Reasoning](https://mlsanity.com/paper/2604.26250/) (2026)

## Abstract (arXiv)

Predicting enzyme kinetic parameters quantifies how efficiently an enzyme catalyzes a specific substrate under defined biochemical conditions. Canonical parameters such as the turnover number (kcat), Michaelis constant (Km), and inhibition constant (Ki) depend jointly on the enzyme sequence, the substrate chemistry, and the conformational adaptation of the active site during binding. Many learning pipelines simplify this process to a static compatibility problem between the enzyme and substrate, fusing their representations through shallow operations and regressing a single value. Such formulations overlook the staged nature of catalysis, which involves both substrate recognition and conformational adaptation. In this regard, we reformulate kinetic prediction as a staged multimodal conditional modeling problem and introduce the Enzyme-Reaction Bridging Adapter (ERBA), which injects cross-modal information via fine-tuning into Protein Language Models (PLMs) while preserving their biochemical priors. ERBA performs conditioning in two stages: Molecular Recognition Cross-Attention (MRCA) first injects substrate information into the enzyme representation to capture specificity; Geometry-aware Mixture-of-Experts (G-MoE) then integrates active-site structure and routes samples to pocket-specialized experts to reflect induced fit. To maintain semantic fidelity, Enzyme-Substrate Distribution Alignment (ESDA) enforces distributional consistency within the PLM manifold in a reproducing kernel Hilbert space. Experiments across three kinetic endpoints and multiple PLM backbones, ERBA delivers consistent gains and stronger out-of-distribution performance compared with sequence-only and shallow-fusion baselines, offering a biologically grounded route to scalable kinetic prediction and a foundation for adding cofactors, mutations, and time-resolved structural cues.