Extracting Latent Sequence-Function Information from Directed Evolution to Design Improved C-H Azidases
9 affiliations
- 1Department of Chemistry
- 2Indiana University
- 3Department of Biochemistry
- 4University of Wisconsin–Madison
- 5Department of Computer Sciences
- 6Morgridge Institute for Research
- 7Department of Biostatistics and Medical Informatics
- 8Department of Biomedical Engineering
- 9Duke University
Abstract
Directed evolution is a powerful approach for enzyme engineering, but conventional workflows use only a fraction of the information generated during experimental screening. Typically, only top-performing variants are sequenced and propagated, while sequence-function information from the remainder of the library is discarded. Here, we develop a data-efficient strategy to recover and leverage this latent information for enzyme design. Screened variants are grouped into activity-defined bins, pooled, and subjected to long-read sequencing, enabling efficient acquisition of sequence-function relationships while preserving relative activity information. Ordinal machine learning models trained on these binned data are then used to predict beneficial mutations. We applied this strategy to SadX, an engineered Fe(II)- and α-ketoglutarate-dependent azidase previously evolved for site-selective C-H azidation. Using only data from a previous directed evolution campaign, the models identified mutations that significantly improved catalytic performance. Experimental validation showed that several predicted single mutations enhanced azidation activity by up to 3-fold relative to the parent enzyme. These results demonstrate that substantial gains in enzyme activity can be achieved by extracting additional information from existing directed evolution datasets, providing a general strategy for converting routinely generated screening data into actionable sequence-function information for enzyme design.
Related
- Directed Evolution of Threonine Aldolases for the Photobiocatalytic Synthesis of β-Branched-α-Tetrasubstituted and Other Noncanonical Amino Acids
- Advances in Machine Learning for Directed Evolution: A Five-Year Retrospective
- Machine learning–assisted directed evolution of plant Rubisco
- PepSpace: An Automated, Physics-Driven Directed Evolution Platform for the De Novo Design of Specific Peptide Binders and Condensate Modulators
- Directed Evolution of a Non-Heme Iron Enzyme for Enantioselective Dearomative Nitrene Transfer: Synthesis of 3a-Aminopyrroloindoline Alkaloids
BibTeX
@misc{ghosh2026extracting,
title = {{Extracting Latent Sequence-Function Information from Directed Evolution to Design Improved C-H Azidases}},
author = {Ayantika Ghosh and Sameer D'Costa and Christian A. Gomez and Bryce Johnson and Prajwal Prabhu and Anthony Gitter and Philip A. Romero and Jared C. Lewis},
howpublished = {ChemRxiv},
year = {2026},
month = oct,
doi = {10.26434/chemrxiv.15009755/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009755/v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Extracting Latent Sequence-Function Information from Directed Evolution to Design Improved C-H Azidases
AU - Ayantika Ghosh
AU - Sameer D'Costa
AU - Christian A. Gomez
AU - Bryce Johnson
AU - Prajwal Prabhu
AU - Anthony Gitter
AU - Philip A. Romero
AU - Jared C. Lewis
PY - 2026
DA - 2026/10/01
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009755/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009755/v1
AB - Directed evolution is a powerful approach for enzyme engineering, but conventional workflows use only a fraction of the information generated during experimental screening. Typically, only top-performing variants are sequenced and propagated, while sequence-function information from the remainder of the library is discarded. Here, we develop a data-efficient strategy to recover and leverage this latent information for enzyme design. Screened variants are grouped into activity-defined bins, pooled, and subjected to long-read sequencing, enabling efficient acquisition of sequence-function relationships while preserving relative activity information. Ordinal machine learning models trained on these binned data are then used to predict beneficial mutations. We applied this strategy to SadX, an engineered Fe(II)- and α-ketoglutarate-dependent azidase previously evolved for site-selective C-H azidation. Using only data from a previous directed evolution campaign, the models identified mutations that significantly improved catalytic performance. Experimental validation showed that several predicted single mutations enhanced azidation activity by up to 3-fold relative to the parent enzyme. These results demonstrate that substantial gains in enzyme activity can be achieved by extracting additional information from existing directed evolution datasets, providing a general strategy for converting routinely generated screening data into actionable sequence-function information for enzyme design.
ER -