KRAS Synthetic Lethal Partners in Cancer

ISEF Category: Biomedical and Health Sciences

Subcategory: Genetics and Molecular Biology of Disease  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

KRAS is one of the most common cancer genes, yet it still resists many drugs. That makes it a huge target for smart data mining. If you can find a gene that KRAS-mutant cells depend on, you may uncover a new weak point in cancer.

What Is It?

Synthetic lethality means two genes become a deadly pair. A cell can survive when either one fails on its own, but if both fail, the cell cannot cope. In cancer research, that idea helps you look for genes that KRAS-mutant cells need more than normal cells do.

Think of KRAS as a broken brake in a car. The cell keeps moving because other systems still work. A synthetic-lethal partner is one of those backup systems. If you find that backup, you may spot a target that matters only in the KRAS-mutant setting, which is the whole point of precision medicine.

This project uses public datasets instead of a wet lab. DepMap gives you gene dependency scores from CRISPR screens, which show what genes different cancer cell lines need to survive. TCGA gives you tumor gene expression data, which helps you see whether a candidate partner moves with KRAS-related patterns across patient samples.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear idea with public data, not just guess from a review article. You connect a real cancer problem, KRAS-driven tumors, to a measurable signal, gene dependency and co-expression. You also get room to ask a fresh question by ranking pairs that have not been reported much. A student can learn data cleaning, correlation, ranking, and basic validation without needing a lab bench.

Research Questions

• How does KRAS mutation status change dependency scores for candidate partner genes in DepMap cell lines?
• What is the effect of filtering by tissue type on the top synthetic-lethal KRAS partner list?
• Does integrating TCGA co-expression improve the ranking of known KRAS-related dependencies?
• To what extent do previously unreported KRAS partner pairs overlap across DepMap and TCGA evidence?
• Which KRAS mutation classes, such as G12D, G12V, or G13D, show the strongest shared dependency patterns?
• How does the choice of correlation cutoff affect the stability of the final candidate set?
• What is the effect of removing low-quality or low-sample cell lines on partner discovery?

Basic Materials

• Laptop or desktop computer with at least 8 GB RAM.
• Stable internet access for downloading public cancer datasets.
• Spreadsheet software for quick sorting and inspection.
• Python installed with pandas, numpy, scipy, and matplotlib.
• R installed with tidyverse and ggplot2, if you prefer R.
• Text editor or notebook for keeping analysis notes.
• Free cloud storage or an external drive for backups.

Advanced Materials

• Workstation with at least 16 GB RAM for larger joins and reshaping.
• Access to Linux or a Unix-like environment for scripted analysis.
• Python with scanpy, seaborn, statsmodels, and scikit-learn.
• R with Bioconductor packages for cancer data handling.
• Cytoscape for network maps of KRAS candidate partners.
• Local mirror of DepMap and TCGA-derived tables for faster iteration.
• Git for version control and reproducible analysis tracking.

Software & Tools

• Python: Cleans the dataset, calculates correlations, and ranks candidate synthetic-lethal pairs.
• RStudio: Helps you make plots, test patterns, and write a clean analysis notebook.
• cBioPortal: Lets you inspect KRAS mutation patterns and expression trends in tumor cohorts.
• DepMap Portal: Gives you CRISPR dependency scores for cancer cell lines.
• Cytoscape: Draws gene interaction networks so you can compare high-priority partners visually.

Experiment Steps

1. Define one KRAS mutation group you will study, then decide whether you are comparing all KRAS-mutant samples or a specific hotspot such as G12D or G12V.
2. Build a candidate gene list from DepMap dependency scores, then decide which threshold or ranking rule will keep the search focused.
3. Match those candidates against TCGA expression patterns, then choose a correlation method that fits the data type.
4. Set up controls that separate KRAS-specific signals from tissue-of-origin effects, batch effects, and noisy low-sample genes.
5. Create a scoring rule that combines dependency, co-expression, and novelty, then decide how you will break ties between similar candidates.
6. Plan one validation check that uses known KRAS synthetic-lethal pairs, then see whether your pipeline can recover them before you trust new hits.

Common Pitfalls

• Mixing cell-line dependency data with tumor expression data without accounting for the difference between model systems.
• Treating every KRAS mutation as identical, which can hide mutation-specific dependency patterns.
• Ranking genes by raw correlation alone, which can elevate tissue markers instead of true partner genes.
• Ignoring sample-size gaps across cancer types, which can make one tissue look stronger than it really is.
• Forgetting to test known KRAS synthetic-lethal pairs first, which makes it hard to trust your pipeline on new pairs.

What Makes This Competitive

A class-level project stops at a short candidate list. A stronger project shows that your pipeline can recover known KRAS partners, survive sensitivity checks, and still surface new pairs. You get an edge by separating tissue effects from mutation effects, then stress-testing your ranking with multiple cutoffs and validation sets. That turns your work from a simple search into a careful discovery pipeline.

Project Variations

• Focus on one KRAS hotspot, such as G12D, and compare its dependency network to other KRAS mutations.
• Swap TCGA co-expression for patient survival links, then see whether the strongest candidates also track with outcomes.
• Add a network analysis step that compares your novel pairs against known pathway neighbors and protein interaction maps.

Learn More

• DepMap Portal: Search the Cancer Dependency Map portal for CRISPR dependency data and tutorials on interpreting gene essentiality.
• Genomic Data Commons: Find TCGA tumor data and metadata through the NCI Genomic Data Commons Data Portal.
• cBioPortal: Explore KRAS mutations, expression trends, and cohort filters in public cancer studies.
• PubMed: Search for review articles on KRAS synthetic lethality and cancer dependency mapping.
• NIH/NCI TCGA Program: Read overview pages on TCGA sample types, sequencing, and data access through the National Cancer Institute.
• MIT OpenCourseWare: Look for free course material on computational biology, statistics, or data analysis if you need a refresher.