Princeton University

• The landmark discovery of penicillin in 1929 ignited a golden era of antibiotic research, revolutionizing medicine and saving countless lives. However, the subsequent decline in the identification of new antibiotics, combined with the alarming rise in antimicrobial resistance, presents an ever-growing public health crisis. In response to this urgent challenge, we present a pioneering approach utilizing a deep neural network, designed to predict molecules with potent antibacterial activity. Leveraging this innovative tool, we conducted extensive predictions on approximately 11 million drug-like compounds meticulously curated from the esteemed ZINC20 database. Our findings reveal an impressive array of candidate molecules, distinguished by their unique structural attributes, promising unprecedented potential in diversifying the antibiotic arsenal against multidrug-resistant pathogens. This work represents a significant step towards addressing the antibiotic crisis and paves the way for a new era of antimicrobial drug discovery.

Penn State

• Ruggedized solvent stable engineered beta-barrel membrane protein-based biomimetic membranes for chemical warfare agent protective fabrics
  • Chemical warfare agents (CWAs) are dangerous weapons of mass destruction that might be used against combatants, first responders, and civilians in terrorist strikes. They have various features, but their main constituents are (1) nerve agents – such as Sarin (Propan-2-oyl methyl phosphonofluoridate) that enter the skin at the rate of 0.1 cm/min and cause suffocation due to lung muscle paralysis, or (2) blister agents such as Mustard gas (MST) (bis-(2-chloroethyl) sulfide) penetrate the skin at the rate of 2.0 cm/min and generate large blisters. Results: We have thus prepared a library of variants of the E. coli-OmpF channel protein that offer a selective bioactive pore interior to bind Sarin and MST using two alternate in silico design routes – IPRO and RosettaDesign. Our results indicate two possible modes of binding enhancement – (1) IPRO designs introduce strong electrostatic attachment with the CWAs, while (2) Rosetta designs are enriched in hydrophobic amino acid substitutions at the CWA binding sites. Both design routes predicted variants that (1) preferentially bind to Sarin – Only Sarin Designs, (2) preferentially bind to MST – Only MST Designs, and (3) Shared Pocket Designs – where two overlapping pockets (with at least one shared non-mutated residue) bind to Sarin and MST respectively.
• Computational Redesign of Acyl-ACP Thioesterase in Umbellularia Californica (UCFatB) with Improved Selectivity Towards Medium-Chain-Length Fatty Acids
  • Enzymes that perform industrially relevant chemistry can provide selectivity otherwise difficult to achieve from chemical synthesis or agriculture. Engineering the acyl-ACP thioesterase in fatty acid biosynthesis has proven an effective approach to tailor the fatty acid distribution toward C₈, C₁₀, and C₁₂ products. To date, the engineering of acyl-ACP thioesterases with a preference for the corresponding medium-chain substrates has been successful with directed evolution and rational mutagenesis techniques. However, a fundamental understanding of the factors driving thioesterase chain-length selectivity remains elusive. Here, we apply a computational method to tune substrate binding and explore the sequence-structure-function relationships of an acyl-ACP thioesterase from Umbellularia californica. We used the Iterative Protein Redesign and Optimization (IPRO) algorithm to computationally design thioesterase variants with improved in silico binding to C₈, C₁₀, and C₁₂ substrates. The pre-screened thioesterase variants were transformed in an Escherichia coli fatty acid production strain to test for changes in fatty acid chain-length distribution from the wild type thioesterase. In this talk, I will describe how we navigated through the design, build, and test cycle to enhance the learnings from each round of mutagenesis.

Sharif University of Technology

• Modeling and Control of an Ethylene Oxide Reactor Using Matlab/Simulink

  •  In this work, the dynamic behavior and control of an industrial fixed-bed ethylene oxide reactor have been investigated. For simulation purposes, a pseudo homogeneous one-dimensional model has been used. First, the reactor simulation is carried out under steady-state conditions and the model is validated by comparison of simulation results with industrial data. Next, a dynamic model is considered. The heat of reaction is used to produce steam and therefore a dynamic model for steam drum is also considered. By manipulating the flow rate of produced steam, the reactor outlet concentration is controlled. Through the dynamic simulation, the process dynamic is approximated by a linear model. This model is used to design a generalized predictive controller (GPC). The performances of the GPC and PI controllers are compared in set-point tracking, load rejection, and model mismatch. The results indicate that the GPC has a better performance. 

• Simulation of a Natural Gas Liquids (NGL) unit in steady-state and dynamic modes using HYSYS software
  • The NGL unit consists of two stages: hydro carbonic liquid separation and refrigeration cycle. The separation part includes five distillation towers and a multi-pass heat exchanger and many shell and tube heat exchangers. First the unit was simulated in steady state then all the equipment was sized and finally the unit was simulated in dynamic mode by using Hysys Dynamic. In this project, the temperature of the last stage of towers, the columns’ top pressure and the level of the condensers and re-boilers are controlled by PID controllers. The performance of PID controllers for load rejection (step input in feed flow) has also been studied.