Title of the talk: "Water and glycerol transport through hydrophobic selectivity
filter: Simulation studies of Trypanasoma brucei aquaglyceroporin channels"

Abstract: "Aquaporins (AQP) and aquaglyceroporins (AQGP) are members of the Major Intrinsic Protein (MIP) Superfamily and they transport water, glycerol and other neutral solutes. They are implicated in several human diseases including autoimmune diseases, neurological disorders, lung infection and cataract. Structures of several MIPs have been determined from different organisms and they all show a unique hourglass helical fold with six transmembrane helices and two half-helices. Our lab has been working on the evolution, structure, function and transport mechanism of several aquaporin homologs. In this talk, I will discuss the ongoing studies in my lab on aquaglyceroporin homolog from a parasite Trypanosoma brucei implicated in African sleeping sickness. Among the three aquaglyceroporin homologs encoded by T. brucei, TbAQP2 is involved in the transport of various substances including drugs that are used to treat the sleeping sickness disease. MIP channels are known to have a narrow constriction region called aromatic/arginine selectivity filter (Ar/R SF) formed by four residues that play vital role in the selectivity and transport of solutes. TbAQP2 is characterized by a unique Ar/R SF formed by all hydrophobic residues. Despite having hydrophobic selectivity filter, TbAQP2 has been shown to transport water, glycerol and other solutes including drug molecules.
We have performed equilibrium molecular dynamics for a period of 1 μs on TbAQP2 embedded in 349 POPE lipids and 29315 water molecules at both sides of the membrane. We analyzed several properties including potential of mean force profiles, average channel radius profiles, osmotic water permeability coefficient, water dipole orientation inside the channels and water residence time inside the channels. All analyzed properties were compared with that of human AQP1, a prototype for water channel. Similarly, umbrella sampling method was used to find out the PMF profiles of glycerol. Our studies revealed the molecular mechanism of water and glycerol transport and explain why TbAQP2 with hydrophobic selectivity filter is more efficient in transporting water and glycerol compared to the prototype water and glycerol channels. This study has implications in developing novel molecules that can interact efficiently with TbAQP2 and will be an alternative to the currently available drugs to treat African sleeping sickness."

Title of the talk: Data-mining and high-throughput approaches to materials discovery

Abstract:

Predicting structure–property correlations for a given composition remains one of the
biggest challenges in materials science[1]. In this talk, I will present our recent progress on
predicting materials using a hybrid materials genome initiative approach—a combined global
minimum search by first principles evolutionary and knowledge based crystal structure
prediction approaches—in searching for the likelihoods of function specific structures. The
methods and results of our investigation on specific important materials will be discussed [2-4].

References

1. P. De Luna, J. Wei, Y. Bengio, A. Aspuru-Guzik, and E. Sargent, Nature 2017, 552 23–25.

2. F. Naaz, M.S. Chauhan, K. Yadav, S. Singh, A. Kumar, and D.L.V.K. Prasad, J. Phys. Chem. C 2024, 128, 4766-4778.

3. F. Naaz and D. L. V. K. Prasad, Bull. Am. Phys. Soc. 2024, B60, 309.

4. H Wickham, J. Stat. Soft. 2014, 59, 1–23; D Adamson et al., arXiv:1910.12026 (2022).

Title of the talk: HPC in the Design of Transport Vehicles: Turboprops, Bullet Trains, and Hyperloops

Abstract: High-Performance Computing (HPC) has become an indispensable tool in accelerating the design and development of modern transport vehicles. Leveraging Computational Fluid Dynamics (CFD) tools on HPC platforms enables engineers to perform realistic, high-fidelity simulations—leading to faster, more cost-effective design cycles. In this talk, we explore the critical role of HPC in the aerodynamic optimization of transport systems, with a focus on turboprops, bullet trains, and Hyperloops. The overarching goal across all platforms is the same: minimize aerodynamic drag to reduce energy consumption. Simulations of such systems pose several challenges, including intricate geometries, highly unsteady forces, and complex flow phenomena. We present results from our in-house simulations that address these complexities. For turboprops, we demonstrate how wing designs can be optimized by accounting for the influence of the propeller slipstream. In the case of bullet trains, our simulations highlight how modifications in the aft-body upsweep region significantly alter the wake flow. For Hyperloop systems, our computational studies provide insights into optimal capsule geometries to meet target performance.

Title of the talk: "Multiscale Materials Modeling via Atomistic, Mesoscale and Machine Learning Based Computations"

Abstract: The role of multiscale materials modeling, integrating atomistic, mesoscale, and machine learning, for comprehensive understanding and predicting the behavior of complex materials will be discussed with specific examples. At the atomistic level, quantum mechanical methods and molecular dynamics simulations capture fundamental interactions and defect behaviors. The mesoscale level bridges microstructural phenomena such as spinodal decomposition and grain growth, which influence macroscopic properties. Machine learning accelerates this multiscale modeling by identifying patterns, reducing computational costs, and impressively enabling rapid predictions of material properties based on vast datasets. This integrated approach provides a robust framework for materials design and optimization, connecting atomic-scale mechanisms with engineering-scale performance more efficiently and accurately.

1. Effect of Cr segregation on grain growth in nanocrystalline α-Fe alloy: A multiscale modeling approach, Computational Materials Science 247, 113509 (2025).

2. Accelerating the prediction of stacking fault energy by combining ab initio calculations and machine learning, Physical Review B 109, 214102 (2024).

3. Accelerating microstructure modelling via machine learning: a new method combining Autoencoder and ConvLSTM, Phys. Rev. Materials 7, 083802 (2023).

4. A robust method of denoising experimental micrographs using deep learning, 223, 114963 (2025).

Title of the talk: Theory-Driven Design of Doped Molten Halide Catalysts

Abstract: Currently, steam reforming of methane (SMR) is the preferred process for the industrial production of hydrogen. Unfortunately, this process generates large quantities of CO2 as an undesirable by-product, which needs in-situ capturing and storage facilities, incurring additional cost to the process. On the other hand, methane pyrolysis to H2 and separable carbon can be used without co-generation of CO2. Recently, in a collaborative effort, we showed that molten catalyst can efficiently convert methane to H2 without coking. In this seminar, I will talk about our recent efforts to develop a new class of catalysts, doped molten salts, for the methane pyrolysis reaction. Through computationally expensive ab initio molecular dynamics (AIMD) simulations, we have discovered that the catalysis in these system proceeds through a very unusual and unexpected mechanism. To access the activity of the doped halide catalyst, we also perform transition state (TS) calculations using the Nudged Elastic Band method (NEB), which involves iteratively optimizing a chain of images connected by artificial springs towards minimum energy path (MEP). A typical NEB calculation in our system has a total of 1000 atoms and approximately 4000 electrons. This constrained optimization calculation requires approximately 5000 steps (or 40000 quantum chemical evaluations) to converge. A calculation on HPC with 4 nodes and 48 cores each (2.9 GHz, 4GB per core) is able to complete 15 steps per hour; requiring 15 days to complete a NEB calculation. To address these issues, we have developed an optimization algorithm, Adaptively Accelerated Relaxation Engine (AARE), which modifies MD trajectories using an adaptive acceleration routine based on local information and a conjugate gradient-like direction. The new optimization routine reduces the number of quantum chemical evaluations by more than 50%. We have also integrated the NEB and AARE with machine learning technique, Gaussian Process Regression (GPR), which decreases the number of quantum chemical evaluations by order-of-magnitude.

Title of the talk: Linear and non-linear Hall effects in quantum materials

Abstract: In this talk, we describe some novel Hall effects in quantum materials, which arise from band geometric quantities in crystalline materials.
Specifically, we will discuss how nonlinear Hall responses, can be used to i) detect topology in the system, and ii) the stacking order.
Finally, we will discuss our efforts to develop the theoretical framework for non-linear spin transport.