Molecular Docking for Covid-19 Therapeutic

As Covid-19 swept across the world, I wanted to do what I could to help. Luckily, a few days later I discovered that my lab (the Al-Hashimi lab) was beginning a brand new project to address the pandemic. Prof. Al-Hashimi is an expert in investigating structural properties of nucleic acids. Nucleic acids (DNA and RNA) are the instruction manuals our cells use to make proteins, and they are vital to cellular function. Inhibiting RNA or DNA could be an incredibly effective therapeutic; however, it is extremely different to find a drug that selectively binds one RNA over the rest. In the last decade of research, he has demonstrated that it is possible to find more specific drugs by considering RNA dynamics and how drugs interact with different conformations of the RNA.

We decided to bring this approach to Covid-19 by using molecular docking (a computer simulation that predicts how well two molecules bind) to find a small molecule drug that binds SARS-CoV2 RNA. Coronavirus’ RNA genomes are actually massive compared to other viruses, providing many potential targets. Our lab decided on the most promising targets, and then generated an ensemble of structures to dock. In this process, I was responsible for two of targets and performed all of the docking for those structures. In addition, I wrote scripts in Python to automate data collection, and I met weekly with Dr. Al-Hashimi and his grad students to review our progress and search for potential therapeutics. We are currently testing the results of the simulation in vitro.

Hours: 100

Supervisor: Prof. Al-Hashimi

Start: April 2020

End: Ongoing

Fast DNA Hybridization Assay

While DNA is usually thought of as a static, linear alpha-helix, its structure actually fluctuates between different conformations. Adopting different conformations is integral to how nucleic acids interact with proteins and other molecules, and these interactions may be manipulated for therapeutic purposes. DNA structure is frequently modified by events like Hoogsten transitions, deaminations, and methylations. We wanted to examine how changing DNA sequence and inducing modifications affect the likelihood of non-Watson-Crick DNA adopting certain dynamic conformations. We hypothesize that these different conformations may explain how modified DNA interacts differently with many proteins. Current methods to monitor changes in DNA structure such as NMR are time-consuming and laborious. The lab has developed a faster approach using UV melting to monitor conformational changes. I have spent the last eight months increasing the throughput of these UV melts from the current standard of 5-10 sequences per day to hundreds of sequences per day.

Genetics labs often use a technique called High Resolution Melting (HRM) to differentiate between DNA strands. It uses special dyes that only fluoresce when bound to double-stranded DNA. HRM heats DNA in the presence of these dyes, and eventually the DNA strands separate, changing fluorescence with temperature. This technique is promising because machines for HRM can run 96 samples per well, increasing throughput 10x with minimal sample preparation. However, in binding to DNA, the dyes change the overall reaction, and it was unknown if the thermodynamic parameters could be measured precisely and accurately.

While at home due to Covid-19 restrictions, I first answered this scientific question by independently building a simulation from scratch in Mathematica. Using data from the literature such as nearest neighbor models and real fluorescence vs temperature curves, I designed rigorous protocols to ensure the model was consistent with previous results. After two months, the model could predict how different DNA sequences and different concentrations of dye or DNA affected the measurement error. Based on these simulations, I identified an ideal ratio of DNA:dye and showed why previous experiments with this technique were unsuccessful. Then, I independently designed an in-vitro experiment to test the model and whether the technique would work. Although the model and experimental design were finished in early June, restrictions from Covid-19 have limited my time in lab, and data collection is still ongoing. I am excited to publish the results early in the Spring Semester.

Hours: 150

Supervisor: Prof. Al-Hashimi

Start: February 2019

End: Ongoing

CPRIT-CURE Summer Internship

In the summer of 2020, I originally was supposed to do research at MD Anderson Cancer Center, but the main research program was dialed down to only 10 hours/week due to Covid-19 restrictions. NevertheIess, I was able to virtually collaborate with Dr. Jia Wu in the imaging physics department on a project that used convolutional neural networks (CNNs) and patient PET/CT data to analyze lung tumor heterogeneity. Dr. Wu had not yet processed data to develop a model; therefore, I worked over twenty hours a week, more than double what was expected, to preprocess scans from over two hundred patients. I attended workshops on programming CNNs, which complemented my experience in the Duke program “Machine Learning in Medicine,” and discussed the intricacies of designing CNNs with Dr. Wu. I am grateful to have collaborated with him because I gained valuable data-processing skills and a new understanding of machine learning’s reliance on massive volumes of organized data. My previous experience with “Machine Learning in Medicine” used a dataset that had already been curated; therefore, I learned a lot about the amount of work that goes into building datasets for machine learning. Going forward, I am certain this exposure to machine learning will help me as I seek to engineer new medicines.

Hours: 200

Supervisor: Dr. Jia Wu, Imaging Physics at MD Anderson Cancer Center

Start: June 2020

End: August 2020