Dr. Nuño, the Chief Scientific Officer at Vapourtec, recently participated in iDMT’s 6th-year gathering, which brought together 70 enthusiastic participants from chemistry, data science, and molecular engineering fields. The conference’s main focus was on exploring cutting-edge research and applications of machine learning, data-driven approaches, and molecular simulations across diverse scientific domains.
Prof. Michael P. Brenner from Harvard University delivered a keynote seminar on harmonising AI software evaluation models, emphasising the use of Chatgpt models in protein structure analysis.
Zsuzsanna Koczor-Benda from the University of Warwick presented her work on “Molecular Screening for Terahertz Detection with Machine-Learning-Based Methods,” introducing a novel approach for predicting spectroscopic properties.
Other notable presentations included Kobi Felton’s “ML-SAFT,” a framework for predicting organic compound thermodynamic properties.
Prof. Pietro Lió’s exploration of generative and graph models in chemistry and medicine, utilising graph neural networks for data analysis.
Timur Madzhidov from Elsevier discussed reaction prediction conditions and challenges related to imbalanced data in software like Reaxys.
Benoît Baillif’s research focused on using atomistic neural networks to bias conformer ensembles for drug targeting.
Tom Savage’s talk on “DARTS” highlighted the impact of 3D printing of reactors on dispersion models.
Meanwhile, in the Vapourtec booth, Dr. Nuño showed the capabilities of the Vapourtec flow chemistry platform RS-400 across a wide range of chemical reactions. Equipping the R-Series with the API package enables the integration of the Vapourtec flow platform with an external machine learning algorithm to monitor, make decisions and new calculations based on the feedback analysis. The R-Series API package includes access to the OPC UA server high and low-level commands.
Dr Nuño demonstrated the two different approaches to externally control the R-Series platform:
Low-level control – Any component of the system can be individually controlled (pumps, valves, reactors’ temperatures, etc) when receiving an external command from a third-party software.
High-level control – After analysing a result, an ML algorithm will decide the next reaction conditions and add a new “reaction” through the R-Series software.
One publication that highlights the benefits is “Accelerated Chemical Reaction Optimization Using Multi-Task Learning” from Prof Alexei Lapkin’s group https://pubs.acs.org/doi/10.1021/acscentsci.3c00050?goto=supporting-info
The iDMT Conference 2023 proved to be a hub of innovation and knowledge exchange, promising significant advancements in the field of molecular engineering and data-driven research.
