Week 8 : Poster Design, Presentations, Peer Learning
This week was all about communication. We focused on designing our research poster for the final symposium, learning how to clearly convey our technical work to a wider audience. We also had discussions on effective presentation skills—how to tell the story of our research, engage the audience, and handle Q&A sessions with confidence. A special highlight was connecting with students from the Math Camp Summer Program, where we shared our research with 7th and 8th graders.
Week 7
This week, we dove deeper into our motion planning pipeline by modifying the basic Probabilistic Roadmap (PRM) algorithm. Instead of sampling the entire configuration space, we defined a user-specified sphere around the active site and sampled only collision-free configurations within that region. After generating the samples, we used the BioPotential distance metric to calculate energy values for each configuration. Finally, we output the energy statistics and began analyzing how our modified method compares with poses that are generated by AutoDock.
Week 6
This week, we reviewed "Sampling-Based Motion Planning for Tracking Evolution of Dynamic Tunnels in Molecular Dynamics Simulations" by Vonasek et al. The paper addresses how to track dynamic protein tunnels over time using sampling-based motion planning. This aligns with our focus on ligand path planning, especially in flexible, moving protein structures. The authors validated their method using real molecular dynamics data to show how tunnels evolve—something static models often miss.
On the coding side, we worked on integrating a new BioPotential distance metric into PPL. We split the class into .h and .cpp files, registered it in the DistanceMetrics list in alphabetical order, and compiled the code using conan, cmake, and makeppl. After fixing build issues, we tested it by setting up the XML configuration and running a sample plan successfully.
Week 5
This week I explored motion planning in the context of protein-ligand interactions. The goal was to simulate how a mobile ligand navigates through a protein structure. Starting with PDB data, I converted the protein into a geometry model and built a flexible ligand. I first generated a random roadmap using an “AlwaysTrue” checker to ignore collisions and visualized it with Vizmo. Next, I created a collision-free roadmap using the “pqp_solid” checker while ignoring adjacent link collisions to improve accuracy. Then, I incorporated potential energy as a distance metric to generate an energetically valid roadmap. There are still some challenges we are facing to implement the metric BioPotential to the PPL library.
Week 4
This week, I focused on understanding motion planning and its implementation using C++. I reviewed a C++ overview to refresh key concepts relevant to robotics programming, then explored a reading assignment on motion planning and its real-world applications, such as autonomous vehicles and robotic navigation. The reading covered fundamental concepts like global vs. local planning and introduced common algorithms like RRT and PRM. I also worked on tasks from the Week 4 folder, which included answering reflection questions and exploring how motion planning is structured within a C++ planning library. This helped solidify my understanding of how theoretical planning concepts are translated into functional code.
Week 3
This week we focused on testing Testing Autodock4 and Caver on 4fwb protein and 3KP ligand. We completed and documentated all the steps required to run the software. Explored user-defined parameters and sensitivity e.g, probe radius, clusterign co-efficient. Also worked on debugging scripts to launch VMD, ended up creating a new script.
Week 2
We began exploring two key tools for predicting protein-ligand binding and accessibility:
CAVER 3.0 – A tool used for tunnel detection in proteins. We learned about its input requirements (PDB structures, atom IDs), output files (CSV, tunnels), and core algorithmic methodology.
AutoDock – A molecular docking tool used to predict binding affinities. We reviewed its input files (.pdbqt for proteins and ligands), docking parameter files, and how it outputs predicted binding modes and energy scores.
The goal was to test both tools using the protein structure 4FWB and ligand 3KP. A big focus has been understanding user-defined parameters in both tools and how sensitive the results are to those choices, something that will be crucial for optimizing predictions and comparing performance with newer tools.
Week 1: Onboarding and Orientation
Completed CITI Program Training on Responsible Conduct of Research
UR2PhD Mentor/Mentee contract