Robert Pollice
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Robert Pollice has a total of 227 co-authors. Here are the top co-authors by number of shared publications:
- Alán Aspuru-Guzik
Department of Chemistry, University of Toronto, Toronto, Canada
19 shared publications - Gabriel dos Passos Gomes
Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, USA
13 shared publications Alan Aspuru-Guzik
University of Toronto
8 shared publications- Peter Chen
Department of Chemistry and Applied Biosciences, Laboratory of Organic Chemistry, ETH Zürich, Zürich 8093, Switzerland
8 shared publications AkshatKumar Nigam
Department of Computer Science, Stanford University, USA
7 shared publications- AkshatKumar Nigam
Department of Computer Science, Stanford University, USA
7 shared publications
Latest Publications
- Open Access
- 8 October 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Active synthesizability prediction for a virtual space of 395 million catalyst candidates with small data
N-heterocyclic carbenes (NHCs) are well established as catalysts and ligands in many chemical transformations. Recently, interest has grown in pyridine[1,5]imidazolium salt pre-catalysts, stable precursors to NHCs, because they are relatively easy to synthesize and highly adaptable. This work introduces a data-driven workflow to identify which pyridine[1,5]imidazolium salts are readily synthetically accessible, starting from exploratory synthesis attempts for 127 compounds. Our approach relies on a feedback loop with a supervised machine learning model to estimate and dynamically update the synthesizability for a combinatorial space of about 394 million pre-catalysts derived from commercially available precursors. Our final model estimates that 4.78% of the entire space are feasible, mapping the space of readily accessible candidates and providing a valuable set for future inverse design studies. We validated the prediction performance of our approach based on both literature data and a holdout set occupying extremes of the combinatorial space.
- 8 October 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessActive synthesizability prediction for a virtual space of 395 million catalyst candidates with small data
N-heterocyclic carbenes (NHCs) are well established as catalysts and ligands in many chemical transformations. Recently, interest has grown in pyridine[1,5]imidazolium salt pre-catalysts, stable precursors to NHCs, because they are relatively easy to synthesize and highly adaptable. This work introduces a data-driven workflow to identify which pyridine[1,5]imidazolium salts are readily synthetically accessible, starting from exploratory synthesis attempts for 127 compounds. Our approach relies on a feedback loop with a supervised machine learning model to estimate and dynamically update the synthesizability for a combinatorial space of about 394 million pre-catalysts derived from commercially available precursors. Our final model estimates that 4.78% of the entire space are feasible, mapping the space of readily accessible candidates and providing a valuable set for future inverse design studies. We validated the prediction performance of our approach based on both literature data and a holdout set occupying extremes of the combinatorial space.
- Open Access
- 16 September 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (9) , 3636-3647
Exhaustive molecular string enumeration for data augmentation and structure exploration
We present TYCHE, an algorithm for the exhaustive enumeration of the molecular string representations SMILES and SELFIES that accounts for all the principal randomization mechanisms.
- 16 September 2026
- Published by Royal Society of Chemistry (RSC) in Digital Discovery
- Vol. 5 (9) , 3636-3647
Open AccessExhaustive molecular string enumeration for data augmentation and structure exploration
We present TYCHE, an algorithm for the exhaustive enumeration of the molecular string representations SMILES and SELFIES that accounts for all the principal randomization mechanisms.
- Open Access
- 25 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Universal Structure Generator: A Simple Plugin Agent For Computational Chemistry and Agentic Frameworks.
Initial structure generation is the universal first step in computational chemistry but remains a bottleneck as each class of structures requires a different structure preparation workflow. We present an open-source UniversalStructure Generator(USG) based on artificial intelligence (AI) that converts natural-language prompts into molecular structures for both molecular and matter simulations, enabling rapid workflow development and seamless agentic AI integration. Computational and theoretical chemistry have recently transitioned into essential parts of modern chemical research. They serve multiple purposes such as supporting experimentally observed behaviour, enabling high-throughput screening before selecting top candidates for synthesis, and revealing underlying mechanisms. Regardless of the research direction or the problem being addressed, the first step in any simulation always remains the same: constructing the initial structure of the system.
- 25 August 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessUniversal Structure Generator: A Simple Plugin Agent For Computational Chemistry and Agentic Frameworks.
Initial structure generation is the universal first step in computational chemistry but remains a bottleneck as each class of structures requires a different structure preparation workflow. We present an open-source UniversalStructure Generator(USG) based on artificial intelligence (AI) that converts natural-language prompts into molecular structures for both molecular and matter simulations, enabling rapid workflow development and seamless agentic AI integration. Computational and theoretical chemistry have recently transitioned into essential parts of modern chemical research. They serve multiple purposes such as supporting experimentally observed behaviour, enabling high-throughput screening before selecting top candidates for synthesis, and revealing underlying mechanisms. Regardless of the research direction or the problem being addressed, the first step in any simulation always remains the same: constructing the initial structure of the system.
- Open Access
- 31 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Universal Structure Generator: A Simple Plugin Agent For Computational Chemistry and Agentic Frameworks.
Initial structure generation is the universal first step in computational chemistry but remains a bottleneck as each class of structures requires a different structure preparation workflow. We present an open-source UniversalStructure Generator(USG) based on artificial intelligence (AI) that converts natural-language prompts into molecular structures for both molecular and matter simulations, enabling rapid workflow development and seamless agentic AI integration. Computational and theoretical chemistry have recently transitioned into essential parts of modern chemical research. They serve multiple purposes such as supporting experimentally observed behaviour, enabling high-throughput screening before selecting top candidates for synthesis, and revealing underlying mechanisms. Regardless of the research direction or the problem being addressed, the first step in any simulation always remains the same: constructing the initial structure of the system.
- 31 July 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessUniversal Structure Generator: A Simple Plugin Agent For Computational Chemistry and Agentic Frameworks.
Initial structure generation is the universal first step in computational chemistry but remains a bottleneck as each class of structures requires a different structure preparation workflow. We present an open-source UniversalStructure Generator(USG) based on artificial intelligence (AI) that converts natural-language prompts into molecular structures for both molecular and matter simulations, enabling rapid workflow development and seamless agentic AI integration. Computational and theoretical chemistry have recently transitioned into essential parts of modern chemical research. They serve multiple purposes such as supporting experimentally observed behaviour, enabling high-throughput screening before selecting top candidates for synthesis, and revealing underlying mechanisms. Regardless of the research direction or the problem being addressed, the first step in any simulation always remains the same: constructing the initial structure of the system.
- Open Access
- 8 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Molecular Simulations Assisted by an Artificial Intelligence Agent (ArIA)
Computational chemistry enables the investigation of molecular and materials properties through simulations. Among available methods, density functional theory (DFT) and time-dependent DFT (TDDFT) are widely applied because they provide practical error margins at affordable computational cost. Despite their broad utility, their effective application requires extensive training in both theoretical chemistry and basic programming, creating a significant entrance barrier for non-specialists. In this work, we develop an agentic artificial intelligence (AI) based on open-source large language models (LLMs) that can be readily deployed on a state-of-the-art personal computers (PCs). This agentic AI is capable of interpreting user instructions, performing simulations with the ORCA program suite, and generating both interactive visualizations and summary reports. Additionally, users can interrogate the simulation results through text queries, making computational chemistry more accessible. We release our agentic AI fully open-source, which will create a collaborative development ecosystem to build upon for future improvements.
- 8 May 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessMolecular Simulations Assisted by an Artificial Intelligence Agent (ArIA)
Computational chemistry enables the investigation of molecular and materials properties through simulations. Among available methods, density functional theory (DFT) and time-dependent DFT (TDDFT) are widely applied because they provide practical error margins at affordable computational cost. Despite their broad utility, their effective application requires extensive training in both theoretical chemistry and basic programming, creating a significant entrance barrier for non-specialists. In this work, we develop an agentic artificial intelligence (AI) based on open-source large language models (LLMs) that can be readily deployed on a state-of-the-art personal computers (PCs). This agentic AI is capable of interpreting user instructions, performing simulations with the ORCA program suite, and generating both interactive visualizations and summary reports. Additionally, users can interrogate the simulation results through text queries, making computational chemistry more accessible. We release our agentic AI fully open-source, which will create a collaborative development ecosystem to build upon for future improvements.
- Open Access
- 23 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Molecular Simulations Assisted by an Artificial Intelligence Agent (ArIA)
Computational chemistry enables the investigation of molecular and materials properties through simulations. Among available methods, density functional theory (DFT) and time-dependent DFT (TDDFT) are widely applied because they provide practical error margins at affordable computational cost. Despite their broad utility, their effective application requires extensive training in both theoretical chemistry and basic programming, creating a significant entrance barrier for non-specialists. In this work, we develop an agentic artificial intelligence (AI) based on open-source large language models (LLMs) that can be readily deployed on a state-of-the-art personal computers (PCs). This agentic AI is capable of interpreting user instructions, performing simulations with the ORCA program suite, and generating both interactive visualizations and summary reports. Additionally, users can interrogate the simulation results through text queries, making computational chemistry more accessible. We release our agentic AI fully open-source, which will create a collaborative development ecosystem to build upon for future improvements.
- 23 April 2026
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessMolecular Simulations Assisted by an Artificial Intelligence Agent (ArIA)
Computational chemistry enables the investigation of molecular and materials properties through simulations. Among available methods, density functional theory (DFT) and time-dependent DFT (TDDFT) are widely applied because they provide practical error margins at affordable computational cost. Despite their broad utility, their effective application requires extensive training in both theoretical chemistry and basic programming, creating a significant entrance barrier for non-specialists. In this work, we develop an agentic artificial intelligence (AI) based on open-source large language models (LLMs) that can be readily deployed on a state-of-the-art personal computers (PCs). This agentic AI is capable of interpreting user instructions, performing simulations with the ORCA program suite, and generating both interactive visualizations and summary reports. Additionally, users can interrogate the simulation results through text queries, making computational chemistry more accessible. We release our agentic AI fully open-source, which will create a collaborative development ecosystem to build upon for future improvements.
- Open Access
Evolutionary exploration of drug-like chemical space utilizing generative AI and virtual screening
The identification of suitable lead molecules in the vast chemical space is a critical and challenging task in drug discovery campaigns. Recently, it has been demonstrated that large-scale virtual screening provides a powerful approach to accelerate the identification of novel drug candidates by screening ever increasing virtual ligand libraries, which have reached magnitudes of > 10 20 compounds. However, this desirable increase in potentially bioactive molecules poses a new challenge as enumerating and virtually screening such huge compound libraries is computationally prohibitive. Consequently, advanced approaches to navigate ultra-large chemical spaces and to identify suitable candidate molecules therein are urgently needed. Here, we present an evolutionary algorithm framework using molecular generative AI, reaction-based substructure searching, and iterative model fine-tuning for a targeted and efficient exploration of chemical fragment spaces. Combining this approach with large-scale virtual screening we are able to identify target-specific candidate molecules within the commercially available Enamine REAL Space (∼10 15 ). We demonstrate the applicability of the approach by successfully identifying and biochemically validating pH-specific ligands of the µ -opioid receptor. Our results demonstrate that integrating generative AI with evolutionary algorithms provides a promising route to explore ultra-large chemical spaces for the discovery of novel, synthetically accessible lead molecules.
Open AccessEvolutionary exploration of drug-like chemical space utilizing generative AI and virtual screening
The identification of suitable lead molecules in the vast chemical space is a critical and challenging task in drug discovery campaigns. Recently, it has been demonstrated that large-scale virtual screening provides a powerful approach to accelerate the identification of novel drug candidates by screening ever increasing virtual ligand libraries, which have reached magnitudes of > 10 20 compounds. However, this desirable increase in potentially bioactive molecules poses a new challenge as enumerating and virtually screening such huge compound libraries is computationally prohibitive. Consequently, advanced approaches to navigate ultra-large chemical spaces and to identify suitable candidate molecules therein are urgently needed. Here, we present an evolutionary algorithm framework using molecular generative AI, reaction-based substructure searching, and iterative model fine-tuning for a targeted and efficient exploration of chemical fragment spaces. Combining this approach with large-scale virtual screening we are able to identify target-specific candidate molecules within the commercially available Enamine REAL Space (∼10 15 ). We demonstrate the applicability of the approach by successfully identifying and biochemically validating pH-specific ligands of the µ -opioid receptor. Our results demonstrate that integrating generative AI with evolutionary algorithms provides a promising route to explore ultra-large chemical spaces for the discovery of novel, synthetically accessible lead molecules.
- 17 November 2025
- Published by American Chemical Society (ACS) in Journal of the American Chemical Society
- Vol. 147 (47) , 43884-43901
Bulky Phosphine Ligands Promote Palladium-Catalyzed Protodeboronation
The Suzuki-Miyaura cross-coupling reaction is plagued by protodeboronation, an undesirable side reaction with water that consumes the boronic acid derivatives required for the cross-coupling reaction. Meticulous mechanistic studies have previously established protodeboronation to be highly sensitive to the nature of the boronic reagent and reaction conditions. Particularly, the presence of bases, which are essential for the Suzuki-Miyaura coupling, is known to catalyze protodeboronation. However, protodeboronation catalyzed by palladium-phosphine complexes, the benchmark catalyst system for Suzuki-Miyaura cross-coupling, has been understudied compared to its base-catalyzed counterpart. We demonstrate, using automated high-throughput experimentation, comprehensive computational mechanistic analyses and kinetic modeling, that protodeboronation is accelerated by palladium(II) complexes bound to bulky phosphine ligands. While sterically hindered ligands are typically used to facilitate difficult cross-couplings, these ligands can instead paradoxically impede cross-coupling product formation, requiring careful and judicious consideration when choosing ligands for Suzuki-Miyaura cross-couplings.
- 17 November 2025
- Published by American Chemical Society (ACS) in Journal of the American Chemical Society
- Vol. 147 (47) , 43884-43901
Bulky Phosphine Ligands Promote Palladium-Catalyzed Protodeboronation
The Suzuki-Miyaura cross-coupling reaction is plagued by protodeboronation, an undesirable side reaction with water that consumes the boronic acid derivatives required for the cross-coupling reaction. Meticulous mechanistic studies have previously established protodeboronation to be highly sensitive to the nature of the boronic reagent and reaction conditions. Particularly, the presence of bases, which are essential for the Suzuki-Miyaura coupling, is known to catalyze protodeboronation. However, protodeboronation catalyzed by palladium-phosphine complexes, the benchmark catalyst system for Suzuki-Miyaura cross-coupling, has been understudied compared to its base-catalyzed counterpart. We demonstrate, using automated high-throughput experimentation, comprehensive computational mechanistic analyses and kinetic modeling, that protodeboronation is accelerated by palladium(II) complexes bound to bulky phosphine ligands. While sterically hindered ligands are typically used to facilitate difficult cross-couplings, these ligands can instead paradoxically impede cross-coupling product formation, requiring careful and judicious consideration when choosing ligands for Suzuki-Miyaura cross-couplings.
- Open Access
- 14 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Uncovering Internal Prediction Mechanisms of Transformer-Based Chemical Foundation Models
Supervised deep learning has become a standard approach to deliver competitive predictive tools that allow relating the structure of molecules and their physicochemical features to properties such as binding to protein targets, performance as electronic materials, and reactivity. However, efforts to understand how these models learn and how specific predictions can be explained are still limited in chemistry, which hinders trust in these predictions. To overcome both the current widespread approach of considering supervised deep learning models as black boxes and the limitations of explainable deep learning based on feature attribution, we introduce two general methods from the field of mechanistic interpretability to molecular property prediction. Specifically, we leverage ablation and adapt existing techniques into the regression lens to inspect model predictions of the Transformer-based ChemBERTa foundation model. Our results for 3 ChemBERTa-based models finetuned on distinct datasets allow us to propose the internal mechanisms operable within each of the corresponding model layers that lead to the final predictions. Our results are a stepping stone towards more trustworthy deep learning models in the molecular domain.
- 14 November 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessUncovering Internal Prediction Mechanisms of Transformer-Based Chemical Foundation Models
Supervised deep learning has become a standard approach to deliver competitive predictive tools that allow relating the structure of molecules and their physicochemical features to properties such as binding to protein targets, performance as electronic materials, and reactivity. However, efforts to understand how these models learn and how specific predictions can be explained are still limited in chemistry, which hinders trust in these predictions. To overcome both the current widespread approach of considering supervised deep learning models as black boxes and the limitations of explainable deep learning based on feature attribution, we introduce two general methods from the field of mechanistic interpretability to molecular property prediction. Specifically, we leverage ablation and adapt existing techniques into the regression lens to inspect model predictions of the Transformer-based ChemBERTa foundation model. Our results for 3 ChemBERTa-based models finetuned on distinct datasets allow us to propose the internal mechanisms operable within each of the corresponding model layers that lead to the final predictions. Our results are a stepping stone towards more trustworthy deep learning models in the molecular domain.
- Open Access
- 19 August 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Bulky phosphine ligands promote palladium-catalyzed protodeboronation
The Suzuki-Miyaura cross-coupling reaction is plagued by protodeboronation, an undesirable side reaction with water that consumes the boronic acid derivatives required for the cross-coupling reaction. Meticulous mechanistic studies have previously established protodeboronation to be highly sensitive to the nature of the boronic reagent and reaction conditions. Particularly, the presence of bases, which are essential for the Suzuki-Miyaura coupling, is known to catalyze protodeboronation. However, protodeboronation catalyzed by palladium-phosphine complexes, the benchmark catalyst system for Suzuki-Miyaura cross-coupling, has been understudied compared to its base-catalyzed counterpart. We demonstrate, using automated high-throughput experimentation, comprehensive computational mechanistic analyses and kinetic modeling, that protodeboronation is accelerated by palladium(II) complexes bound to bulky phosphine ligands. While sterically hindered ligands are typically used to facilitate difficult cross-couplings, these ligands can instead paradoxically impede cross-coupling product formation, requiring careful and judicious consideration when choosing ligands for Suzuki-Miyaura cross-couplings.
- 19 August 2025
- Published by American Chemical Society (ACS) in ChemRxiv
Open AccessBulky phosphine ligands promote palladium-catalyzed protodeboronation
The Suzuki-Miyaura cross-coupling reaction is plagued by protodeboronation, an undesirable side reaction with water that consumes the boronic acid derivatives required for the cross-coupling reaction. Meticulous mechanistic studies have previously established protodeboronation to be highly sensitive to the nature of the boronic reagent and reaction conditions. Particularly, the presence of bases, which are essential for the Suzuki-Miyaura coupling, is known to catalyze protodeboronation. However, protodeboronation catalyzed by palladium-phosphine complexes, the benchmark catalyst system for Suzuki-Miyaura cross-coupling, has been understudied compared to its base-catalyzed counterpart. We demonstrate, using automated high-throughput experimentation, comprehensive computational mechanistic analyses and kinetic modeling, that protodeboronation is accelerated by palladium(II) complexes bound to bulky phosphine ligands. While sterically hindered ligands are typically used to facilitate difficult cross-couplings, these ligands can instead paradoxically impede cross-coupling product formation, requiring careful and judicious consideration when choosing ligands for Suzuki-Miyaura cross-couplings.