Drug Discovery From Your Laptop
Pharma spends millions on this exact process. You can do the first step for free and turn it into a real, publishable paper.
Pharma spends millions on this exact process. You can do the first step for free, and turn it into a real, publishable paper.
The Idea
Most diseases are driven by specific proteins in your body. If you can find a small molecule that fits into that protein and blocks it, you have a potential drug candidate. That single idea is the foundation of modern medicine — pharmaceutical companies spend millions running this exact pipeline, and the first, computational step of it is something you can do on a laptop, for free.
It’s called molecular docking (virtual screening). You take a target protein, test thousands of chemical compounds against it, and score how well each one binds. This isn’t a toy — the exact workflow below (AlphaFold structures + AutoDock Vina docking) is used in real, published drug-discovery research, and your version is a genuinely publishable output.
Your Free Toolkit
Four free tools do the whole job. Set them up before you start.
- AlphaFold Protein Structure Database. Built by Google DeepMind + EMBL-EBI. A free database with the predicted 3D structure of nearly every known protein — over 200 million of them. Download the 3D structure of your target protein and pair it with UniProt to find which protein belongs to your disease.
- AutoDock Vina. Built by Scripps Research (free & open-source). The docking engine that places each compound into the protein and scores how well it binds. Run your compounds against the protein, get candidate molecules free from PubChem or ZINC, and view results in UCSF Chimera or PyMOL.
- Claude. Built by Anthropic. An AI assistant for the science itself — not a docking program, but the brain beside you. Ask it which compounds to try and why, have it write and debug your docking scripts, and get it to interpret your scores and flag likely false positives.
- Instudi’s AI Counselor. Built by Instudi (free support). Your admissions-and-research guide for when you get stuck. Ask it how to structure the project, what to do when a step breaks, and how to turn your results into a paper and an application asset.
The Full Step-by-Step Breakdown
Choose a disease with a known protein driver — usually a specific enzyme or receptor. On UniProt, search the disease or gene, find the associated human protein, and note its ID.
From the protein’s page, open its AlphaFold structure and download the PDB file. If an experimental structure exists in the Protein Data Bank, prefer it — experimental structures dock more reliably than predictions.
Convert everything into the format AutoDock Vina needs (PDBQT) using free tools like AutoDockTools, Meeko, or Open Babel, and define a search box around the protein’s binding site. This part is fiddly — have Claude write and debug the prep scripts with you.
Dock each compound against the protein with AutoDock Vina. Every compound gets a binding-affinity score in kcal/mol. Loop over your whole library to screen hundreds or thousands of compounds automatically.
The more negative the score, the stronger the predicted binding — so the lowest scores are your best matches. Sort your compounds, take the top hits as your candidate drug molecules, and view their poses in Chimera or PyMOL to confirm they actually sit in the pocket. Ask Claude to help interpret the results.
Frame it around a clear hypothesis (e.g., “compounds of class X will bind protein Y better than known inhibitor Z”), then report your method, your ranked candidates, and your limits. That structure is what makes it publishable — and it’s the next section.
Turn It Into a Publication
Your ranked candidates and clean method are a real, publishable output. Here’s where high schoolers actually publish this kind of work — for free.
Journal of Emerging Investigators (JEI)
A free, open-access, peer-reviewed journal founded by Harvard graduate students in 2011, built specifically for middle and high school researchers. It’s free to submit and free to publish, uses genuine two-stage peer review (graduate-student and faculty reviewers), and publishes students from 20+ countries. Two things to plan for: you’ll need an adult mentor (a teacher, parent, or advisor) to submit, and your paper must be framed as a hypothesis-driven study. Find the guidelines at emerginginvestigators.org.
Before (or alongside) a journal: post it publicly. Put a preprint and a public GitHub repo online with your pipeline, data, and results. It’s timestamped, citable, and a link an admissions officer can click today — while peer review takes a few months.
Why It Wins — and What’s Next
Almost no high schooler has done real computational drug discovery, and even fewer have published it. It’s specific, verifiable, and shows exactly the self-directed technical curiosity top pre-med and bio programs look for. The next steps get harder — molecular-dynamics simulation to test whether a hit stays bound, machine-learning rescoring for accuracy, and ADMET checks for whether a compound could ever be a real drug. Eventually, email a professor to see whether a top hit is worth validating in a real lab.
Stuck on your project? Ask Instudi’s AI counselor.
Instudi is the college admissions platform that turns a project like this into an application that gets in: free support, real frameworks, and the strategy for the next four years.