Behrgen Smith
(bear-gin)
Data Scientist | Computational Chemistry
Dedicated to leveraging computer science to push forward the next generation of drug discovery. I strive to find novel approaches to solving technical problems. I utilize technical experience in many aspects of drug discovery including ADMET modeling, molecular dynamics, and applications of machine learning to understand mechanisms of pathogenicity.
I pride myself on my ability to mix 3D animation with my research into molecular systems to engage my audiences and communicate the dynamics of molecular systems effectively.
Unless otherwise noted, all videos and graphics on this site were created myself.
EXPERIENCE
Developed classical machine learning models including support vector machines, random forests, and decision tree classifiers to model ADMET endpoints for hit-to-lead optimization. Additionally performed maintenance and added tools to our drug discovery platform.
May 2022 - May 2023
Congruence Therapeutics
Computational Chemistry Intern
University of Pittsburgh - TECBIO REU
NSF Research Intern
Served as intern at the Bahar Lab in the Department of Computational and Systems Biology. In my time at the lab I assisted on two docking projects and developed automated protocols to speed up analysis completing both projects several weeks ahead of deadline. Additionally I aided in projects related to evaluating the impact of mutations in the intrinsic dynamics of proteins and how this information can be leveraged for drug design.
May 2023 - Current
Congruence Therapeutics
Data Scientist I
Currently working in the Computational Chemistry Group. Lead developer on inclusion of membrane effects into our platform. Additionally, perform support roles in development of web applications, as well as machine learning, and data analytics to support our discovery phase research and the lead optimization performed by the medicinal chemistry department.
May 2021-Aug 2021
Currivulum Vitae
Attached is an overview of both my academic and professional experience, including my research interests
EDUCATION
Bachelor's Degree
2019-2023
ENGINEERING SCHOOL
Milwaukee School of Engineering
Milwaukee, WI
Major : Biomolecular & Chemical Engineering
Minor : Mathematics
GPA : 3.79
PROJECTS
2020-2021
My first major 3D animation and first time leading a
team in a structural biology investigation.
In 2020-2021, my sophomore year of college, I joined a club called CREST (Connecting Researchers Educators and Students). This would be our intro to structural biology and reading research papers. During the following months we would learn about different types of intermolecular forces and how they translate to the design and efficacy of small and large molecule therapeutics.
Since this occurred during COVID quarantine, I utilized the extra time it provided to teach myself 3D animation with the purpose of augmenting the presentation of results from our investigation.
As valuable as I found my experience in CREST in understanding the structural determinants of therapeutics, the program unfortunately lost funding at the end of that school year, as those managing the grant funding wanted to focus on the high school counterpart. With MSOE being a primarily engineering school and the Biomolecular Engineering program I was a part of being a small program, there were very few opportunities for undergraduate research. The loss of CREST was one of the last programs remaining.
However, using the knowledge I gained from my internship that coming summer, I sought out and recruited research advisors, Dr. David Koes at the University of Pittsburgh and Dr. Chris Cunningham at the Concordia Wisconsin University to support the formation of a new CREST the following academic year. The previous iteration only allowed for basic structural investigations, however with the tools I learned at TECBIO, I created a curriculum to teach the basics of drug discovery at the undergraduate level. We focused on virtual screenings of Mur-E ligase and used it as a case system to investigate how the intrinsic dynamics of proteins can be used to guide drug discovery as well as evaluating the changes in ligand binding affinity as the protein samples different conformations. Every week I would prepare a lesson and would teach it to 14-20 peers, with the top 4 participators coming to present our research at 2 conference trips I led.
I am proud to say that several people (all seen below) who eventually obtained research internships have told me that those who hired them said their participation in CREST and the research skills they learned from it were a key part of their admission. I strive to learn and succeed, but also lift others up as I go. In my time leading CREST, I learned how challenging it can be to start something new from scratch, to have others count on you for insight and technical ability, and the challenge of creating a consistently engaging and understandable research environment for learning.
DiscoverBMB Conference
Seattle, 2023
ASBMB Conference
Philadelphia, 2022
ASBMB Conference
Philadelphia, 2022
In 2021 I worked under Dr. Ivet Bahar at the University of Pittsburgh's Department of Computational and Systems Biology as an intern in the TECBIO REU. During my time I used docking simulations via AutoDock and AutoDock Vina to attempt to elucidate the mechanism of action for allosteric modulators of the EAAT2 Glutamate Transporter. We also used coarse grained modeling techniques to understand the elevator-like dynamics of the transporter and evaluate how the intrinsic dynamics of the protein change upon ligand binding.
My novel contributions include developing an automated protocol for deploying docking simulation that enabled us to complete the project 3 weeks ahead of time and work on another project which would publish results later that year.
I used what I learned in TECBIO to start a drug discovery club at my university, successfully I leveraged the accessibility of computational biology research to enable undergraduate research at a university where little occurred and was recognized for my efforts in an alumni feature by TECBIO.
(Link to Article Below)
An example of the dynamics of the transporter we investigated. The radial motion of the transporter is reconfigured in the presence of the membrane into a vertical, elevator-like motion. The red domain is the "scaffold" which remains rigid, while the blue "elevator" domain, moves beside it. By sampling different conformations and evaluate how the binding affinity of the allosteric modulators changed in different locations, we could determine a probable molecular basis for their effect.
Publication from TECBIO
Romanazzi T, Zanella D, Cheng MH, Smith B, Carter AM, Galli A, Bahar I, Bossi E. Bile Acids Gate Dopamine Transporter Mediated Currents. Front Chem. 2021 Dec 10;9:753990. doi: 10.3389/fchem.2021.753990. PMID: 34957043; PMCID: PMC8702627.
My current role is as a Data Scientist I at Congruence Therapeutic's Computational Chemistry Group. One of my larger roles this year has been the inclusion of membrane effects into our drug design workflows. This has involved overcoming many technical challenges including making custom changes to the source code of OpenMM and performing many experiments to fine tune the molecular dynamics protocol used. These challenges arose from having to reconcile the company's novel, propriety tools for drug design that operate on a different paradigm than classical molecular dynamics.
Montreal. QC
Quantitative Structure Activity Relationships of Small Molecules
Decision Trees /
Random Forests
Support Vector Machines via
Forward Feature Selection
-
Started with classical machine learning to model ADMET endpoints
-
Quick turnover times
-
Generally explainable and interpretable
-
Developed procedural workflows for generating local models efficiently with applications in ADMET properties and predicting modes of pathogenicity.
-
-
Developed custom visualization to map internal experiments to commercial machine learning models, both of which produce missing values. This necessitated creating a correlation matrix can show the statistics for each pairing independently.
-
This project was used to advise leadership as to whether to join a molecular modeling partnership.
-
Y-axis represents Experimental Endpoints
X-axis represents related machine learning models
The first number of each entry is the correlation, followed by the p-value, and finally the sample size, that can vary by row and column.
Posters
CCG Users Group Meeting - 2023 Montreal, QC
DiscoverBMB Conference - 2023 Seattle, WA
ASBMB Conference - 2022 Philadelphia, PA
CCG Users Group Meeting - 2023 Montreal, QC
EXPERTISE
I have several years of experience studying the intrinsic dynamics of proteins, from allosteric modulators to antibiotics. I have experience in working with source code of molecular dynamics programs as well as performing parameterization to ensure realistic conformations are sampled
I have advanced skills in manipulating data in a variety of formats, from chemical structures to experimental data and vectors of motion for biological systems. When possible, I prefer to use analytical means to reach data driven solutions to maximize explainability, transparency, and minimize complexity.
Molecular Dynamics
Data Analytics
Diverse Background
While most of my current work is in computational chemistry. I do formal training in wet-lab work and have diverse array of experience in PCR, Cell Culture, and other applications of biotechnology. Combined with a degree in Chemical Engineering and a solid background in computer science, I believe I can tackle a wide range of problems.