Seminars

Coarse Grained Force Field Optimization using Machine Learning Methods

Date: Nov. 9, 2023, 2:11 p.m.


Seminar Details: The development of accurate and efficient force fields is essential for the simulation of molecular processes in computational biophysics and materials science. Coarse-grained (CG) models offer a promising approach by reducing the complexity of molecular systems, allowing for the investigation of larger systems and longer timescales. However, the simplification inherent in CG models necessitates careful parameterization to ensure that essential physical properties are retained. Traditional force field optimization relies heavily on iterative, manual adjustments based on intuition and comparison with experimental or high-level theoretical data, which can be time-consuming and may not capture the multidimensional nature of the parameter space. This presentation will explore the integration of machine learning (ML) methods into the optimization of CG force fields. We will discuss how ML algorithms can be employed to systematically improve force field parameters by learning from a diverse set of training data, including structural, thermodynamic, and dynamic properties of molecules. The focus will be on the application of Genetic Algorithm, to predict the quality of force field parameters and guide the optimization process.

Details

Dr. Yesudasan is an Assistant Professor of Mechanical Engineering at the University of New Haven, with a rich background that encompasses a decade of expertise split between academia and industry. He holds a PhD in Mechanical and Aerospace Engineering from Syracuse University. His research is multifaceted, focusing on the development of advanced force fields for water, the intricacies of radiative heat transfer, and the complex modeling of blood clots. With five years of prior professorship and an equal span in the industrial sector, Dr. Yesudasan brings a balanced perspective to his teaching and a practical edge to his innovative research endeavors.



Upcoming Seminars


Date Speaker or Topic Invitee Location Notes
Nov. 11, 2022, 7:11 p.m. Ilsa Cooke, Ph.D Zoom: https://unewhaven.zoom.us/j/99564249101 Time: 11:00 AM – 12:15 PM Location: Kaplan Hall 107
Nov. 4, 2022, 7:11 p.m. James R. Cheeseman, Ph.D. Zoom: https://unewhaven.zoom.us/j/99564249101 Location: Kaplan Hall 107
Oct. 28, 2022, 7:10 p.m. Chris Shingledecker, Ph. D. Zoom: https://unewhaven.zoom.us/j/99564249101 Location: Kaplan Hall 107
Sept. 8, 2023, 11:09 a.m. Dr. Stacy Malaker https://unewhaven.zoom.us/j/97804977329 BCST [Bergami] 307 (Peterson Auditorium)
Sept. 21, 2023, 11:09 a.m. Pak Kin Wong, PhD https://unewhaven.zoom.us/j/97804977329 BCST [Bergami] 307 (Peterson Auditorium)
Aug. 10, 2023, 3:08 a.m. Dr. Yongsheng Chen on-ground University of New Haven
Oct. 20, 2023, 11:10 a.m. Alicia O. Hernandez-Castillo , PhD https://unewhaven.zoom.us/j/97804977329 October 17, 2023 (Friday) Location: BCST [Bergami] 307 (Peterson Auditorium)
Nov. 9, 2023, 2:11 p.m. Sumith Yesudasan Daisy Ph.D. Univ. of New Haven, Main Campus/Virtual Option Host: Biomedical Engineering Graduate Program, Univ. Of New Haven Where: Univ. of New Haven, Main Campus/Virtual Option On 11/9 12:30pm-2:00pm on Thursday. Organizing Team: Kagya Amoako, PhD; kamoako@newhaven.edu Hyesun Cheong; Hcheong@newhaven.edu
Dec. 1, 2023, 11:12 a.m. Dr. Dan Maser BCST [Bergami] 307 (Peterson Auditorium) https://unewhaven.zoom.us/j/97804977329


Seminar Details


Less than a hundred years ago astronomers believed that molecules could not survive in the harsh environment of interstellar space. However, advancements in radio astronomy in the last 50 years have enabled a boom in the detection of new molecules. Today, our picture of the molecular universe has expanded and over 250 molecules have been identified in the interstellar medium, including exotic and unstable species as well as many organic molecules that are also found on Earth. These molecules can be used to probe the temperatures, densities and levels of radiation and ionization of interstellar regions. Interstellar space also represents the ultimate physical chemistry laboratory, providing the ideal testing grounds for fundamental theories of cold chemistry and collisions. The reactions of molecules in interstellar space occur over a wide range of temperatures, down to 10 K or colder, in both the gas-phase and on the surface of icy dust grains. This presents an enormous challenge for laboratory astrochemists, who aim to simulate conditions in space and quantify formation routes of interstellar molecules. I will discuss experimental methods that can be used to simulate these environments and support astronomical observations, including ice experiments and low-temperature gas-phase reaction kinetics measurements using the CRESU (French acronym for Reaction Kinetics in Uniform Supersonic Flow) technique. I will present the major challenges in the field and how we use experiments to approach these questions.

Chiroptical spectroscopy is a form of spectroscopy in which one measures the difference in response of a chiral molecule between left and right circularly polarized light or radiation. Chiroptical spectroscopic techniques are widely used, especially in the pharmaceutical industry, as they play a fundamental role in determining the absolute configuration of chiral molecules. Most forms of chiroptical spectroscopy can now be simulated using quantum chemistry programs and this aspect has become an integral part of these techniques. In this seminar, I will present a brief overview of the different forms of chiroptical spectroscopy along with a few applications. Additionally, I will share my perspective as a quantum chemist.

Since the middle of the last century, our knowledge of interstellar molecules has revolutionized our understanding of the "molecular universe". Beyond the most common molecule, H2, we now know of more than 200 distinct species, not counting many isotopologues. It even appears that the building blocks of biochemical molecules, proteins and DNA/RNA might be formed in interstellar space and could trigger the development of life on developing planets. In this talk, I give an overview of our knowledge of astrochemistry, and highlight some of my recent research in the field, most recently as part of one of the first groups to use the James Webb Space Telescope.

Mucin-domain glycoproteins are densely O-glycosylated and play key roles in a host of biological functions. However, their dense O-glycosylation remains enigmatic both in glycoproteomic landscape and structural dynamics, primarily due to the challenges associated with studying mucin domains. Here, we present advances in the mass spectrometric analysis of mucins, including the characterization of mucinases, software comparisons, and complete mucinomic mapping of translationally relevant mucin proteins.

Recent advances in molecular engineering and advanced imaging techniques have enabled the analysis of complex biological systems with unprecedented throughput and resolution. These novel biosensing techniques, when combined with microfluidic devices and artificial intelligence-guided workflows, offer exciting opportunities for the next generation of precision health. However, current biosensing techniques often require physical isolation or cell lysis, leading to the loss of important phenotypic features, such as drug resistance, invasiveness, and inflammatory responses. To address this unmet need in precision medicine, we are developing dynamic single cell biosensors and ex vivo disease models for rapid antimicrobial susceptibility testing, point-of-care stone metabolic workup, and treatment optimization with patient-derived tumor organoids. In this presentation, I will discuss the application of these technologies for precision management of urological diseases, including the rapid diagnosis of urinary tract infections caused by multidrug-resistant bacteria and the clinical management of muscle invasive bladder cancer.

The disinfection of drinking water using chlorine-based methods was a significant public health achievement in the 20th century, greatly reducing the risk of waterborne diseases caused by microbes. However, today’s chlorinated drinking water still poses safety concerns due to the presence of trace amounts of regulated and unregulated disinfection byproducts (DBPs), as well as other known, unknown, and emerging contaminants (KUECs) that can pose chronic risks and need to be removed. Conventional chemical-based drinking water treatment processes are not effective at removing DBPs or KUECs, and thus alternative approaches are necessary to minimize these risks by targeting the removal of DBP precursors and KUECs that are commonly found in water supplies. In this talk, we introduce the "Minus Approach," a novel approach to water treatment that mitigates KUECs and DBPs without compromising the safety of microbiological quality. The Minus Approach aims to produce biologically stable water with minimal human health risk and significantly lower concentrations of KUECs and DBPs, while reducing the use of chemical treatments that may cause problems, in contrast to the conventional "Plus Approach." By avoiding primary chemical-based coagulants, disinfectants, and advanced oxidation processes, the Minus Approach avoids primary chemical-based coagulants, disinfectants, and advanced oxidation processes, and instead focuses on membrane-based physical separation techniques that can effectively remove DBP precursors and pathogens from the main water treatment system, minimize KUECs generation in the finished water. If needed, the concentrate can be treated separately. Furthermore, the Minus Approach integrates with artificial intelligence to optimize and enhance its performance. Implementing the Minus Approach in water treatment can lead to improved sustainability by reducing reliance on chemical treatments and minimizing harmful contaminants in treated water.

To understand a molecule’s behavior, we first need to determine its structure and how it interacts with its surroundings. Microwave spectroscopy is an excellent tool to unambiguously determine the structure of molecules. Rotational spectra are sensitive to the bond lengths and bond angles present in the molecules and to the positions and orientations of molecules that are bound together via intermolecular forces. The molecules presented in this talk all have in common a succinimide moiety, a five membered cycle in which there are three polar groups (two carbonyl groups and an amide group in the middle). The broadband rotational spectra of these molecules have been acquired and best-fit rotational parameters have been determined for the molecules in this series. The newly constructed experimental set-up at Harvey-Mudd College, the details of the data acquisition, and the future directions of our research will be discussed.

The development of accurate and efficient force fields is essential for the simulation of molecular processes in computational biophysics and materials science. Coarse-grained (CG) models offer a promising approach by reducing the complexity of molecular systems, allowing for the investigation of larger systems and longer timescales. However, the simplification inherent in CG models necessitates careful parameterization to ensure that essential physical properties are retained. Traditional force field optimization relies heavily on iterative, manual adjustments based on intuition and comparison with experimental or high-level theoretical data, which can be time-consuming and may not capture the multidimensional nature of the parameter space. This presentation will explore the integration of machine learning (ML) methods into the optimization of CG force fields. We will discuss how ML algorithms can be employed to systematically improve force field parameters by learning from a diverse set of training data, including structural, thermodynamic, and dynamic properties of molecules. The focus will be on the application of Genetic Algorithm, to predict the quality of force field parameters and guide the optimization process.

The optical frequency comb is a technological wonder of optical physics, earning its creators the Nobel Prize in Physics in 2005, and has only recently become accessible to scientists outside of the world’s premier research institutions. The optical frequency comb is a broadband laser that is essentially the combination of hundreds of thousands of continuous-wave lasers. Frequency combs have proven to be a powerful tool in precision molecular spectroscopy due to their unique blend of a coherent, low-noise spectral source and the broad bandwidth they offer. One particularly useful technique to take full advantage of the high-resolution potential of these devices is dual-comb spectroscopy, in which two frequency combs with slightly different repetition rates are mixed together to generate an interferogram, akin to the signal an FTIR produces, through which the optical spectrum can be retrieved.In this seminar, I will provide an overview of the theory, design, and applications, of the optical frequency comb, discuss our construction of the pair currently in operation, and outline our work towards stabilizing the combs for their upcoming use in precision molecular spectroscopy