Jeevan Tewari
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jeevan Tewari has a total of 23 co-authors. Here are the top co-authors by number of shared publications:
- Aashish Manglik
Department of Pharmaceutical Chemistry, University of California, San Francisco, San Francisco, CA, USA
4 shared publications - Christian Bache Billesbølle
Department of Pharmaceutical Chemistry, University of California, San Francisco, USA
4 shared publications - Claire A. de March
Department of Molecular Genetics and Microbiology, Duke University, Durham, USA
4 shared publications - Hiroaki Matsunami
Duke University School of Medicine
4 shared publications - Nagarajan Vaidehi
Department of Computational and Quantitative Medicine, Beckman Research Institute of the City of Hope, Duarte, CA, USA
4 shared publications Ning Ma
Beckman Research Institute of the City of Hope
3 shared publications
Latest Publications
- Open Access
- 31 July 2026
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 15
Benchmarking biochemical networks generated by large language models
Computational models of biochemical networks provide frameworks for predicting how molecular cues guide cell decisions. These models are typically limited by the time-intensive manual curation required to extract network mechanisms from incomplete literature. Here, we test whether general-purpose large language models (LLMs) can generate accurate models of signaling and metabolic networks. We find that general-purpose LLMs generate 24–65% of the reactions of literature-curated signaling networks for cardiomyocyte hypertrophy, myofibroblast activation, and mechanosignaling. Further, logic-based models based on these networks predict responses to perturbations with accuracies of 6–33%. In the context of metabolic modeling, LLMs are able to generate 64–91% of the reactions within the core Escherichia coli metabolic network and demonstrate highly variable accuracies in predicting substrate utilization. Current general-purpose LLMs generate biochemical networks with moderate accuracy, and this study provides a pipeline and benchmarks to guide future improvements.
- 31 July 2026
- Published by eLife Sciences Publications, Ltd in eLife
- Vol. 15
Open AccessBenchmarking biochemical networks generated by large language models
Computational models of biochemical networks provide frameworks for predicting how molecular cues guide cell decisions. These models are typically limited by the time-intensive manual curation required to extract network mechanisms from incomplete literature. Here, we test whether general-purpose large language models (LLMs) can generate accurate models of signaling and metabolic networks. We find that general-purpose LLMs generate 24–65% of the reactions of literature-curated signaling networks for cardiomyocyte hypertrophy, myofibroblast activation, and mechanosignaling. Further, logic-based models based on these networks predict responses to perturbations with accuracies of 6–33%. In the context of metabolic modeling, LLMs are able to generate 64–91% of the reactions within the core Escherichia coli metabolic network and demonstrate highly variable accuracies in predicting substrate utilization. Current general-purpose LLMs generate biochemical networks with moderate accuracy, and this study provides a pipeline and benchmarks to guide future improvements.
- Open Access
Benchmarking biochemical networks generated by large language models
Computational models of biochemical networks provide frameworks for predicting how molecular cues guide cell decisions. These models are typically limited by the time-intensive manual curation required to extract network mechanisms from incomplete literature. Here, we test whether general-purpose large language models (LLMs) can generate accurate models of signaling and metabolic networks. We find that general-purpose LLMs generate 24-65% of the reactions of literature-curated signaling networks for cardiomyocyte hypertrophy, myofibroblast activation, and mechanosignaling. Further, logic-based models based on these networks predict responses to perturbations with accuracies of 6-33%. In the context of metabolic modeling, LLMs are able to generate 64-91% of the reactions within the core Escherichia coli metabolic network and demonstrate highly variable accuracies in predicting substrate utilization. Current general-purpose LLMs generate biochemical networks with moderate accuracy, and this study provides a pipeline and benchmarks to guide future improvements.
Open AccessBenchmarking biochemical networks generated by large language models
Computational models of biochemical networks provide frameworks for predicting how molecular cues guide cell decisions. These models are typically limited by the time-intensive manual curation required to extract network mechanisms from incomplete literature. Here, we test whether general-purpose large language models (LLMs) can generate accurate models of signaling and metabolic networks. We find that general-purpose LLMs generate 24-65% of the reactions of literature-curated signaling networks for cardiomyocyte hypertrophy, myofibroblast activation, and mechanosignaling. Further, logic-based models based on these networks predict responses to perturbations with accuracies of 6-33%. In the context of metabolic modeling, LLMs are able to generate 64-91% of the reactions within the core Escherichia coli metabolic network and demonstrate highly variable accuracies in predicting substrate utilization. Current general-purpose LLMs generate biochemical networks with moderate accuracy, and this study provides a pipeline and benchmarks to guide future improvements.
- 30 October 2024
- Published by Springer Nature in Nature
- Vol. 635 (8038) , 499-508
Engineered odorant receptors illuminate the basis of odour discrimination
How the olfactory system detects and distinguishes odorants with diverse physicochemical properties and molecular configurations remains poorly understood. Vertebrate animals perceive odours through G protein-coupled odorant receptors (ORs)1. In humans, around 400 ORs enable the sense of smell. The OR family comprises two main classes: class I ORs are tuned to carboxylic acids whereas class II ORs, which represent most of the human repertoire, respond to a wide variety of odorants2. A fundamental challenge in understanding olfaction is the inability to visualize odorant binding to ORs. Here we uncover molecular properties of odorant–OR interactions by using engineered ORs crafted using a consensus protein design strategy3. Because such consensus ORs (consORs) are derived from the 17 major subfamilies of human ORs, they provide a template for modelling individual native ORs with high sequence and structural homology. The biochemical tractability of consORs enabled the determination of four cryogenic electron microscopy structures of distinct consORs with specific ligand recognition properties. The structure of a class I consOR, consOR51, showed high structural similarity to the native human receptor OR51E2 and generated a homology model of a related member of the human OR51 family with high predictive power. Structures of three class II consORs revealed distinct modes of odorant-binding and activation mechanisms between class I and class II ORs. Thus, the structures of consORs lay the groundwork for understanding molecular recognition of odorants by the OR superfamily.
- 30 October 2024
- Published by Springer Nature in Nature
- Vol. 635 (8038) , 499-508
Engineered odorant receptors illuminate the basis of odour discrimination
How the olfactory system detects and distinguishes odorants with diverse physicochemical properties and molecular configurations remains poorly understood. Vertebrate animals perceive odours through G protein-coupled odorant receptors (ORs)1. In humans, around 400 ORs enable the sense of smell. The OR family comprises two main classes: class I ORs are tuned to carboxylic acids whereas class II ORs, which represent most of the human repertoire, respond to a wide variety of odorants2. A fundamental challenge in understanding olfaction is the inability to visualize odorant binding to ORs. Here we uncover molecular properties of odorant–OR interactions by using engineered ORs crafted using a consensus protein design strategy3. Because such consensus ORs (consORs) are derived from the 17 major subfamilies of human ORs, they provide a template for modelling individual native ORs with high sequence and structural homology. The biochemical tractability of consORs enabled the determination of four cryogenic electron microscopy structures of distinct consORs with specific ligand recognition properties. The structure of a class I consOR, consOR51, showed high structural similarity to the native human receptor OR51E2 and generated a homology model of a related member of the human OR51 family with high predictive power. Structures of three class II consORs revealed distinct modes of odorant-binding and activation mechanisms between class I and class II ORs. Thus, the structures of consORs lay the groundwork for understanding molecular recognition of odorants by the OR superfamily.
- Open Access
Engineered odorant receptors illuminate structural principles of odor discrimination
A central challenge in olfaction is understanding how the olfactory system detects and distinguishes odorants with diverse physicochemical properties and molecular configurations. Vertebrate animals perceive odors via G protein-coupled odorant receptors (ORs). In humans, ∼400 ORs enable the sense of smell. The OR family is composed of two major classes: Class I ORs are tuned to carboxylic acids while Class II ORs, representing the vast majority of the human repertoire, respond to a wide variety of odorants. How ORs recognize chemically diverse odorants remains poorly understood. A fundamental bottleneck is the inability to visualize odorant binding to ORs. Here, we uncover fundamental molecular properties of odorant-OR interactions by employing engineered ORs crafted using a consensus protein design strategy. Because such consensus ORs (consORs) are derived from the 17 major subfamilies of human ORs, they provide a template for modeling individual native ORs with high sequence and structural homology. The biochemical tractability of consORs enabled four cryoEM structures of distinct consORs with unique ligand recognition properties. The structure of a Class I consOR, consOR51, showed high structural similarity to the native human receptor OR51E2 and yielded a homology model of a related member of the human OR51 family with high predictive power. Structures of three Class II consORs revealed distinct modes of odorant-binding and activation mechanisms between Class I and Class II ORs. Thus, the structures of consORs lay the groundwork for understanding molecular recognition of odorants by the OR superfamily.
Open AccessEngineered odorant receptors illuminate structural principles of odor discrimination
A central challenge in olfaction is understanding how the olfactory system detects and distinguishes odorants with diverse physicochemical properties and molecular configurations. Vertebrate animals perceive odors via G protein-coupled odorant receptors (ORs). In humans, ∼400 ORs enable the sense of smell. The OR family is composed of two major classes: Class I ORs are tuned to carboxylic acids while Class II ORs, representing the vast majority of the human repertoire, respond to a wide variety of odorants. How ORs recognize chemically diverse odorants remains poorly understood. A fundamental bottleneck is the inability to visualize odorant binding to ORs. Here, we uncover fundamental molecular properties of odorant-OR interactions by employing engineered ORs crafted using a consensus protein design strategy. Because such consensus ORs (consORs) are derived from the 17 major subfamilies of human ORs, they provide a template for modeling individual native ORs with high sequence and structural homology. The biochemical tractability of consORs enabled four cryoEM structures of distinct consORs with unique ligand recognition properties. The structure of a Class I consOR, consOR51, showed high structural similarity to the native human receptor OR51E2 and yielded a homology model of a related member of the human OR51 family with high predictive power. Structures of three Class II consORs revealed distinct modes of odorant-binding and activation mechanisms between Class I and Class II ORs. Thus, the structures of consORs lay the groundwork for understanding molecular recognition of odorants by the OR superfamily.
- Top Cited Papers
- 15 March 2023
- Published by Springer Nature in Nature
- Vol. 615 (7953) , 742-749
Structural basis of odorant recognition by a human odorant receptor
Our sense of smell enables us to navigate a vast space of chemically diverse odour molecules. This task is accomplished by the combinatorial activation of approximately 400 odorant G protein-coupled receptors encoded in the human genome1,2,3. How odorants are recognized by odorant receptors remains unclear. Here we provide mechanistic insight into how an odorant binds to a human odorant receptor. Using cryo-electron microscopy, we determined the structure of the active human odorant receptor OR51E2 bound to the fatty acid propionate. Propionate is bound within an occluded pocket in OR51E2 and makes specific contacts critical to receptor activation. Mutation of the odorant-binding pocket in OR51E2 alters the recognition spectrum for fatty acids of varying chain length, suggesting that odorant selectivity is controlled by tight packing interactions between an odorant and an odorant receptor. Molecular dynamics simulations demonstrate that propionate-induced conformational changes in extracellular loop 3 activate OR51E2. Together, our studies provide a high-resolution view of chemical recognition of an odorant by a vertebrate odorant receptor, providing insight into how this large family of G protein-coupled receptors enables our olfactory sense.
- 15 March 2023
- Published by Springer Nature in Nature
- Vol. 615 (7953) , 742-749
Top Cited PapersStructural basis of odorant recognition by a human odorant receptor
Our sense of smell enables us to navigate a vast space of chemically diverse odour molecules. This task is accomplished by the combinatorial activation of approximately 400 odorant G protein-coupled receptors encoded in the human genome1,2,3. How odorants are recognized by odorant receptors remains unclear. Here we provide mechanistic insight into how an odorant binds to a human odorant receptor. Using cryo-electron microscopy, we determined the structure of the active human odorant receptor OR51E2 bound to the fatty acid propionate. Propionate is bound within an occluded pocket in OR51E2 and makes specific contacts critical to receptor activation. Mutation of the odorant-binding pocket in OR51E2 alters the recognition spectrum for fatty acids of varying chain length, suggesting that odorant selectivity is controlled by tight packing interactions between an odorant and an odorant receptor. Molecular dynamics simulations demonstrate that propionate-induced conformational changes in extracellular loop 3 activate OR51E2. Together, our studies provide a high-resolution view of chemical recognition of an odorant by a vertebrate odorant receptor, providing insight into how this large family of G protein-coupled receptors enables our olfactory sense.
- Open Access
Structural basis of odorant recognition by a human odorant receptor
Our sense of smell enables us to navigate a vast space of chemically diverse odor molecules. This task is accomplished by the combinatorial activation of approximately 400 olfactory G protein-coupled receptors (GPCRs) encoded in the human genome 1–3 . How odorants are recognized by olfactory receptors (ORs) remains mysterious. Here we provide mechanistic insight into how an odorant binds a human olfactory receptor. Using cryogenic electron microscopy (cryo-EM), we determined the structure of active human OR51E2 bound to the fatty acid propionate. Propionate is bound within an occluded pocket in OR51E2 and makes specific contacts critical to receptor activation. Mutation of the odorant binding pocket in OR51E2 alters the recognition spectrum for fatty acids of varying chain length, suggesting that odorant selectivity is controlled by tight packing interactions between an odorant and an olfactory receptor. Molecular dynamics simulations demonstrate propionate-induced conformational changes in extracellular loop 3 to activate OR51E2. Together, our studies provide a high-resolution view of chemical recognition of an odorant by a vertebrate OR, providing insight into how this large family of GPCRs enables our olfactory sense.
Open AccessStructural basis of odorant recognition by a human odorant receptor
Our sense of smell enables us to navigate a vast space of chemically diverse odor molecules. This task is accomplished by the combinatorial activation of approximately 400 olfactory G protein-coupled receptors (GPCRs) encoded in the human genome 1–3 . How odorants are recognized by olfactory receptors (ORs) remains mysterious. Here we provide mechanistic insight into how an odorant binds a human olfactory receptor. Using cryogenic electron microscopy (cryo-EM), we determined the structure of active human OR51E2 bound to the fatty acid propionate. Propionate is bound within an occluded pocket in OR51E2 and makes specific contacts critical to receptor activation. Mutation of the odorant binding pocket in OR51E2 alters the recognition spectrum for fatty acids of varying chain length, suggesting that odorant selectivity is controlled by tight packing interactions between an odorant and an olfactory receptor. Molecular dynamics simulations demonstrate propionate-induced conformational changes in extracellular loop 3 to activate OR51E2. Together, our studies provide a high-resolution view of chemical recognition of an odorant by a vertebrate OR, providing insight into how this large family of GPCRs enables our olfactory sense.