Parkinson’s Drug Repurposing With Gene Signatures

ISEF Category: Cellular and Molecular Biology

Subcategory: Neurobiology  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

Parkinson's disease slowly changes how brain cells work long before symptoms get severe. That means the problem leaves a molecular trail you can search for in public data. With the right tools, you can compare that trail to drugs already known to affect genes. You are not guessing, you are matching patterns.

What Is It?

This project starts with a simple idea. Diseases change gene activity. Drugs change gene activity too. If a disease creates a harmful gene pattern, you can look for drugs that push that pattern in the opposite direction. That is called signature reversal.

Think of it like noise-canceling headphones. The disease creates one sound pattern, and you want a second pattern that cancels it out. Connectivity-Map, also called CMap, and LINCS L1000 give you large public datasets that record how cells respond to many compounds. You can compare a Parkinson's-related gene signature from substantia nigra RNA-seq data to those drug-response signatures.

After you get a short list of candidate drugs, you can rank them again using AlphaFold-based structural clues. AlphaFold predicts protein shapes, and binding pockets are places where a drug might fit. This does not prove a drug will work, but it helps you ask which hits look more plausible before you move to lab validation.

Why This Is a Good Topic

This is a strong science fair topic because it gives you a real research workflow with public data, a clear output, and room for original analysis. You can test whether known or overlooked drugs reverse Parkinson's signatures, then compare those hits with structural druggability clues. The topic connects to a major medical problem, and you can learn bioinformatics, statistics, and target prioritization without starting with a wet lab.

Research Questions

• How does the disease gene signature differ across public substantia nigra RNA-seq datasets??
• What is the effect of using different differential expression cutoffs on the number of drug hits that reverse the Parkinson's signature??
• Does combining multiple substantia nigra datasets improve the stability of the top Connectivity-Map candidates??
• To what extent do the highest-ranked drug hits overlap with drugs already linked to dopamine or alpha-synuclein biology??
• Which top-ranked compounds also score well for binding-pocket plausibility in AlphaFold-predicted target structures??
• How does the choice of reference gene set change the drug reversal ranking??

Basic Materials

• Laptop or desktop computer with at least 16 GB RAM.
• Stable internet connection for downloading public datasets and using web tools.
• Spreadsheet software for organizing hit lists and scores.
• R or Python installed for differential expression, ranking, and plotting.
• PubMed access for checking prior evidence on candidate compounds.
• NIH GEO or ArrayExpress access for public RNA-seq datasets.
• Connectivity-Map web access for searching compound signatures.
• AlphaFold Protein Structure Database access for predicted protein structures.
• ImageJ or similar image viewer for inspecting plots and heatmaps.

Advanced Materials

• Access to a Linux workstation or cloud compute instance for larger RNA-seq processing.
• RStudio or Jupyter Notebook for reproducible analysis.
• Bioconductor packages for RNA-seq normalization and differential expression.
• Molecular docking software for follow-up binding analysis, if your mentor approves it.
• Protein structure visualization software such as PyMOL or UCSF ChimeraX.
• Curated Parkinson's gene sets from public databases or review articles.
• CSV exports from LINCS L1000 or Connectivity-Map query results.
• Version control system such as Git for tracking analysis changes.

Software & Tools

• R: Runs differential expression, enrichment, and ranking analyses for RNA-seq data.
• Python: Helps clean candidate lists, score overlaps, and make reproducible plots.
• GEOquery: Pulls public transcriptomics datasets from the NIH GEO database into R.
• Enrichr: Checks whether your top genes cluster in pathways linked to Parkinson's biology.
• PyMOL: Lets you inspect AlphaFold protein models and inspect likely binding pockets.

Experiment Steps

1. Define the exact Parkinson's comparison you will test, such as one substantia nigra dataset or a merged set of public datasets.
2. Build the disease gene signature from public RNA-seq data and decide how you will rank up-regulated and down-regulated genes.
3. Match that signature against Connectivity-Map or LINCS L1000 drug-response profiles and decide how you will score reversal strength.
4. Filter the top compound list by evidence quality, known mechanism, and overlap with Parkinson's pathways.
5. Compare the best hits against AlphaFold protein models and decide which targets have believable binding-pocket features.
6. Plan a validation strategy that uses independent datasets, pathway enrichment, or literature-backed target checks to support your top candidate.

Common Pitfalls

• Using one noisy RNA-seq dataset as if it represents all Parkinson's biology, which makes the signature too fragile.
• Mixing gene identifiers from different databases without careful mapping, which breaks the Connectivity-Map query.
• Ranking drugs only by reversal score and ignoring whether the target is actually expressed in substantia nigra.
• Treating AlphaFold structure alone as proof of druggability, which overstates what a predicted model can tell you.
• Forgetting to test whether the top hits stay similar when you change the cutoff, dataset, or gene ranking method.

What Makes This Competitive

A class-level project usually stops at one query and one ranked list. A stronger project tests whether the top hits stay stable across datasets, ranking methods, and pathway filters. You can also earn points by comparing multiple evidence layers, such as reversal score, target expression, and structural plausibility. That kind of cross-checking makes your pipeline look much more like real translational research.

Project Variations

• Use a different neurodegenerative dataset, such as substantia nigra data from a second cohort, and compare whether the same drugs rise to the top.
• Focus on a smaller target class, such as dopamine-linked genes or alpha-synuclein interactors, and see whether the ranking becomes more specific.
• Swap the structural filter for pathway enrichment only, then compare whether the two prioritization methods point to the same compounds.

Learn More

• NCBI GEO: Search for public substantia nigra RNA-seq datasets and download expression data.
• PubMed: Search for review articles on Parkinson's transcriptomics and drug repurposing.
• NIH LINCS Program: Read about Connectivity-Map and L1000 drug-response signatures.
• AlphaFold Protein Structure Database: Look up predicted protein structures and inspect candidate targets.
• MIT OpenCourseWare: Search for free lectures on genomics, bioinformatics, and systems biology.
• Nature Reviews Neurology: Search for review articles on Parkinson's molecular biology and therapeutic targets.