Dual-graph deconvolution of enzyme fitness: an interpretable framework to identify and explain catalytic modulators
Models trained on protein sequences can predict mutational effects, but their outputs represent a convolution of different phenotypes. Isolating stability-independent activity effects is essential for rational enzyme design, allosteric drug design and functional annotation, yet existing pLM-based deconvolution methods lack interpretability. Here, we introduce a hybrid framework that pairs…
An innovative framework combines stability predictors with probabilistic graphical models of coevolution to isolate the effects of enzyme mutations that do not affect stability. This hybrid approach, trained on protein sequence models, accurately predicts the independent activity effects of mutations in enzymes. Benchmarked against enzymes with available multidimensional deep mutational scanning data, the framework matches the performance of existing probabilistic language model-based methods.
Moreover, the researchers have identified biophysically plausible communication pathways that align with existing knowledge and validated these findings experimentally across a variety of enzyme targets. By merging statistical and geometric deep learning techniques, this zero-shot tool offers an interpretable method for targeted engineering and mechanistic discovery in enzyme design and allosteric drug development.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
