EnzymeCAGE: A Geometric Foundation Model for Enzyme Retrieval with Evolutionary Insights DOI Creative Commons
Yong Liu,

Chenqing Hua,

Tao Zeng

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Abstract Enzyme catalysis is fundamental to life, driving the chemical transformations that sustain biological processes and support industrial applications. However, unraveling intertwined relationships between enzymes their catalytic reactions remains a significant challenge. Here, we present EnzymeCAGE, catalytic-specific geometric foundation model trained on approximately 1 million structure-informed enzyme-reaction pairs, spanning over 2,000 species encompassing an extensive diversity of genomic metabolic information. EnzymeCAGE features geometry-aware multi-modal architecture coupled with evolutionary information integration module, enabling it effectively nuanced enzyme structure, function, reaction specificity. supports both experimental predicted structures applicable across diverse families, accommodating broad range metabolites types. Extensive evaluations demonstrate EnzymeCAGE’s state-of-the-art performance in function prediction, de-orphaning, site identification, biosynthetic pathway reconstruction. These results highlight its potential as transformative for understanding accelerating discovery novel biocatalysts.

Language: Английский

Microbial Technologies Enhanced by Artificial Intelligence for Healthcare Applications DOI Creative Commons

Taeho Yu,

Minjee Chae,

Ziling Wang

et al.

Microbial Biotechnology, Journal Year: 2025, Volume and Issue: 18(3)

Published: March 1, 2025

ABSTRACT The combination of artificial intelligence (AI) with microbial technology marks the start a major transformation, improving applications throughout biotechnology, especially in healthcare. With capability AI to process vast amounts biological big data, advanced allows for comprehensive understanding complex systems, advancing disease diagnosis, treatment and development therapeutics. This mini review explores impact AI‐integrated technologies healthcare, highlighting advancements biomarker‐based therapeutics production therapeutic compounds. exploration promises significant improvements design implementation health‐related solutions, steering new era biotechnological applications.

Language: Английский

Citations

0
Chemoenzymatic synthesis planning by evaluating the synthetic potential in biocatalysis and chemocatalysis DOI Creative Commons
Xuan Liu, Hongxiang Li, Huimin Zhao

et al.

Published: Aug. 30, 2024

Chemoenzymatic synthesis integrates the advantages of chemocatalysis and biocatalysis to design efficient routes. However, current computer-assisted chemoenzymatic planning tools lack a heuristic method unify step-by-step molecule-by-molecule identification chemo-/biocatalysis opportunities in Here we develop an asynchronous retrosynthesis algorithm (ACERetro) which employs search strategy that prioritizes exploration molecule's promising catalytic methods. The suitability molecule be synthesized via chemo- or is quantitatively evaluated by data-driven Synthetic Potential Score (SPScore) using neural network model. Additionally, SPScore can used heuristically identify For given route, this uses molecules offer optimization potential when alternative method, then ACERetro Case studies on for ethambutol epidiolex demonstrate our concise routes applying enzymatic steps introduce stereochemistry find shortcuts. Moreover, case route rivastigmine (R,R)-formoterol how finds bypasses form alternative, shorter Our findings with evaluating synthetic represents versatile effective framework planning.

Language: Английский

Citations

1
Computer-Aided Retrosynthesis for Greener and Optimal Total Synthesis of a Helicase-Primase Inhibitor Active Pharmaceutical Ingredient DOI Creative Commons

Rodolfo I. Teixeira,

Michael Andresini, Renzo Luisi

et al.

JACS Au, Journal Year: 2024, Volume and Issue: 4(11), P. 4263 - 4272

Published: Oct. 2, 2024

This study leverages and upgrades the capabilities of computer-aided retrosynthesis (CAR) in systematic development greener more efficient total synthetic routes for active pharmaceutical ingredient (API) IM-204, a helicase-primase inhibitor that demonstrated enhanced efficacy against Herpes simplex virus (HSV) infections. Using various CAR tools, several were uncovered, evaluated, experimentally validated, with goal to maximize selectivity yield minimize environmental impact. The tools revealed options under different constraints, which can overperform patented route used as reference. selected CAR-based significant improvement from 8% (patented route) 26%, along moderate overall green performance. It was also shown human-in-the-loop approach be synergistically combined drive further improvements deliver alternatives. strategy metrics by substituting solvents merging two steps into one. These changes led IM-204 synthesis 8 35%. Additionally, performance score, based on GreenMotion metrics, improved 0 18, cost building blocks reduced 550-fold. work demonstrates potential drug development, highlighting its capacity streamline processes, reduce footprint, lower production costs, thereby advancing field toward sustainable practices.

Language: Английский

Citations

1
Yeast-MetaTwin for Systematically Exploring Yeast Metabolism through Retrobiosynthesis and Deep Learning DOI Creative Commons
Ke Wu, Haohao Liu,

Manda Sun

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 2, 2024

Abstract Underground metabolism plays a crucial role in understanding enzyme promiscuity, cellular metabolism, and biological evolution, yet experimental exploration of underground is often sparse. Even though yeast genome-scale metabolic models have been reconstructed curated for over 20 years, more than 90% the metabolome still not covered by these models. To address this gap, we developed workflow based on retrobiosynthesis deep learning methods to comprehensively explore metabolism. We integrated predicted network into consensus model, Yeast8, reconstruct twin Yeast-MetaTwin, covering 16,244 metabolites (92% total metabolome), 2,057 genes 59,914 reactions. revealed that K m parameters differ between known network, identified hub molecules connecting pinpointed percentages pathways. Moreover, Yeast-MetaTwin can predict by-products chemicals produced yeast, offering valuable insights guide engineering designs.

Language: Английский

Citations

0
Chemoenzymatic Multistep Retrosynthesis with Transformer Loops DOI Creative Commons
David Kreutter, Jean‐Louis Reymond

Chemical Science, Journal Year: 2024, Volume and Issue: unknown

Published: Jan. 1, 2024

Integrating enzymatic reactions into computer-aided synthesis planning (CASP) should help devise more selective, economical, and greener synthetic routes.

Language: Английский

Citations

0
Chemoenzymatic Synthesis Planning Guided by Reaction Type Score DOI Creative Commons
Hongxiang Li, X. Liu, Guangde Jiang

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 29, 2024

ABSTRACT Thanks to the growing interests in computer-aided synthesis planning (CASP), a wide variety of retrosynthesis and retrobiosynthesis tools have been developed past decades. However, for multi-step chemoenzymatic reactions are still rare despite widespread use enzymatic chemical synthesis. Herein we report reaction type score (RTscore)-guided (RTS-CESP) strategy. Briefly, RTscore is trained using text-based convolutional neural network (TextCNN) distinguish from decomposition evaluate efficiency. Once multiple routes generated by tool target molecule, used rank them find step(s) that can be replaced improve As proof concept, RTS-CESP was applied 10 molecules with known literature able predict all six being top-ranked routes. Moreover, employed 1000 boutique database 554 molecules, outperforming ASKCOS, state-of-the-art tool. Finally, design new route FDA-approved drug Alclofenac, which shorter than literature-reported experimentally validated.

Language: Английский

Citations

0
Chemoenzymatic Synthesis Planning Guided by Reaction Type Score DOI
Hongxiang Li, Xuan Liu, Guangde Jiang

et al.

Journal of Chemical Information and Modeling, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 8, 2024

Thanks to the growing interest in computer-aided synthesis planning (CASP), a wide variety of retrosynthesis and retrobiosynthesis tools have been developed past decades. However, for multistep chemoenzymatic reactions are still rare despite widespread use enzymatic chemical synthesis. Herein, we report reaction type score (RTscore)-guided (RTS-CESP) strategy. Briefly, RTscore is trained using text-based convolutional neural network (TextCNN) distinguish from decomposition evaluate efficiency. Once multiple routes generated by tool target molecule, used rank them find step(s) that can be replaced improve As proof concept, RTS-CESP was applied 10 molecules with known literature able predict all six being top-ranked routes. Moreover, employed 1000 boutique database 554 molecules, outperforming ASKCOS, state-of-the-art tool. Finally, design new route FDA-approved drug Alclofenac, which shorter than literature-reported has experimentally validated.

Language: Английский

Citations

0
EnzymeCAGE: A Geometric Foundation Model for Enzyme Retrieval with Evolutionary Insights DOI Creative Commons
Yong Liu,

Chenqing Hua,

Tao Zeng

et al.

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Abstract Enzyme catalysis is fundamental to life, driving the chemical transformations that sustain biological processes and support industrial applications. However, unraveling intertwined relationships between enzymes their catalytic reactions remains a significant challenge. Here, we present EnzymeCAGE, catalytic-specific geometric foundation model trained on approximately 1 million structure-informed enzyme-reaction pairs, spanning over 2,000 species encompassing an extensive diversity of genomic metabolic information. EnzymeCAGE features geometry-aware multi-modal architecture coupled with evolutionary information integration module, enabling it effectively nuanced enzyme structure, function, reaction specificity. supports both experimental predicted structures applicable across diverse families, accommodating broad range metabolites types. Extensive evaluations demonstrate EnzymeCAGE’s state-of-the-art performance in function prediction, de-orphaning, site identification, biosynthetic pathway reconstruction. These results highlight its potential as transformative for understanding accelerating discovery novel biocatalysts.

Language: Английский

Citations

0