Cancer Research SocietySeven Lakes HS
01 / 20
Plus: how to start your own independent research.
Cancer isn't one disease or one mutation. It's a Darwinian process: a single cell picks up a mutation, divides, and its descendants pick up more. Cells that happen to grow faster or die less out-compete their neighbors and take over.
That's why tumors are genetically messy and why they evolve resistance to drugs — you're not fighting one enemy, you're fighting a population that keeps mutating. Every division is another roll of the dice.
One mutation is never enough. Cancer accumulates hits in stages — initiation (first mutation), promotion (those cells expand), progression (more mutations, invasion). The textbook case is the colon:
Ordinary colon cells.
The first brake fails.
Growth accelerates.
Now invasive cancer.
Why risk climbs with age — and why a colonoscopy that removes a polyp prevents a cancer.
A dividing cell moves through four phases — G1 (grow), S (copy the DNA), G2 (check the copy), M (split). Cyclins and CDKs drive it forward like gears.
At each checkpoint, the cell asks: is the DNA intact? Big enough? Correctly copied? If not, it pauses to repair — or triggers self-destruction. The restriction point in G1 is the point of no return.
Cancer = these checkpoints stop saying "no."
DNA is copied billions of times, and damage from UV, tobacco, radiation, and copy errors piles up. Mutations come in flavors: point mutations, insertions/deletions, copy-number changes, and translocations — like the Philadelphia chromosome (BCR-ABL) behind CML, shut down by imatinib.
Only a few driver mutations cause cancer; the rest are passengers. Repair systems normally fix errors — when they break (BRCA1/2), damage accumulates. The twist: BRCA-mutant tumors die from PARP inhibitors (“synthetic lethality”).
Oncogenes are the accelerator. A proto-oncogene (RAS, MYC, HER2, EGFR) mutates into overdrive — and it only takes one broken copy to floor it. Gain-of-function.
Tumor suppressors are the brakes (TP53, RB, PTEN, BRCA). Here you must lose both copies — Knudson's "two-hit" rule. Loss-of-function. TP53, the "guardian of the genome," is mutated in over half of all cancers.
Insight: you need both — gas and failed brakes — to crash.
↑ one hit floors the gas · both brakes must fail
Growth signals travel through molecular relay chains. Cancer mutates one node to weld the switch on — and every node is a drug target.
Relays “divide.” Mutant RAS (~1 in 3 cancers) or BRAF jams it on. Example: BRAF+MEK inhibitors in melanoma.
Relays “grow & survive.” The brake PTEN is often lost; targeted by PI3K and mTOR inhibitors.
EGFR mutations drive many lung cancers — and EGFR inhibitors target them. (CRS's EGFR L858R project lives right here.)
However a cancer starts, it converges on the same set of acquired capabilities. Master these ten and you understand almost every therapy ever designed.
Apoptosis is clean, programmed suicide — the cell packages itself into neat fragments, no mess. Two triggers converge on executioner caspases: the intrinsic path (internal damage → mitochondria, gated by the BCL-2 / BAX balance) and the extrinsic path (external death signals).
Cancer tips that balance toward survival — often by overexpressing BCL-2. Example: venetoclax, a BCL-2 inhibitor, restores apoptosis in leukemia (CLL).
Your chromosomes end in protective caps called telomeres. They shorten a little with every division — a built-in counter. When they run out, a normal cell stops dividing (senescence).
Around 90% of cancers switch the enzyme telomerase back on, rebuilding the caps so the counter never hits zero. That's the difference between a cell that divides ~50 times and one that divides forever.
No cell can live more than ~1–2 mm from a blood vessel — oxygen won't diffuse further. So a tumor can't grow large… until it flips the "angiogenic switch."
It secretes VEGF, chemically begging nearby vessels to sprout new branches toward it. The new vessels are leaky and chaotic — which is exactly what helps cells escape later. This is why anti-VEGF drugs (bevacizumab) try to starve tumors.
Metastasis — not the primary tumor — causes ~90% of cancer deaths. Cells undergo EMT (they loosen their grip on neighbors), break into a blood vessel, survive the bloodstream, squeeze out elsewhere, and try to colonize a foreign organ. It's staggeringly inefficient — fewer than 0.01% succeed — which is exactly why it's so hard to stop the few that do.
A tumor isn't just cancer cells — it's a whole microenvironment of blood vessels, immune cells, and support cells the tumor recruits and corrupts.
Cancer cells burn glucose by glycolysis even when oxygen is plentiful — inefficient for energy, but fast, and it supplies building blocks for new cells. It's why PET scans (which track glucose uptake) light tumors up.
T-cells should kill cancer. Tumors hide by displaying PD-L1, a molecular "don't-eat-me" flag. Checkpoint inhibitors (pembrolizumab) rip that flag off — the breakthrough behind modern immunotherapy.
Tumor-associated fibroblasts and macrophages are tricked into supplying growth signals, remodeling tissue, and even shielding cancer from drugs.
How abnormal the cells look under the microscope. Low grade = near-normal, slow. High grade = wild, disorganized, fast. It’s about the cells themselves.
How far it has spread. T = tumor size/invasion, N = lymph nodes, M = distant metastasis — combined into Stage I–IV.
Stage drives treatment and prognosis more than tumor type. T1N0M0 (Stage I) and T3N2M1 (Stage IV) are worlds apart.
“What kind” matters less than “how far” — catching it early is everything.
You don't need to be in a lab coat to do it. Research is asking a specific, unanswered question and producing evidence. Four doors are open to you right now:
Docking, modeling, ML — real oncology research on your laptop. The most accessible path, and CRS's specialty.
Public patient datasets (TCGA, SEER) hold thousands of unanswered questions. Just needs a spreadsheet and a hypothesis.
Bench work through a mentor — or a literature review, a legitimate first paper that needs zero equipment.
You need a good question, a method, and a mentor — not a million-dollar lab.
Pick a topic you like. Read reviews on PubMed & Google Scholar until you can explain it.
Watch for "future work," "remains unknown," "poorly understood." That's your opening.
Feasible, Interesting, Novel, Ethical, Relevant — the test of a good question.
"Does drug X bind kinase Y better than the known inhibitor?" beats "cure cancer."
Never read top to bottom. Read in this order — you'll grasp a paper in 15 minutes instead of two hours:
The abstract tells you if it matters; the figures are the results. Understand the figures and you understand the paper.
The last lines of the intro state the actual question. The discussion admits the limitations — often your next project.
Read methods closely just for the technique you plan to reuse. Tools: PubMed, PMC (free full text), Connected Papers, preprints (bioRxiv).
Short. Specific. Proves you read their work. Clear, small ask. Send 30, not 3.
Computational oncology needs no wet lab — just a question and free tools. This is exactly how CRS members ran the DCLK1 and EGFR projects.
Predict how a drug fits a protein pocket. Grab a structure from the Protein Data Bank, dock with AutoDock Vina, visualize in ChimeraX.
Predict activity or classify tumors with Python + scikit-learn. Re-score structures with Boltz-2. Free, and it ends in a manuscript.
Python basics, then follow one published pipeline end-to-end. Our masterclasses walk you through each.
An under-studied cancer protein — say the DCLK1 kinase. Download its 3-D structure from the PDB (7F3G).
Gather compounds to test — e.g. every FDA-approved drug (thousands, free).
Clean the protein, define the binding pocket, set up ligands in AutoDock Tools.
AutoDock Vina predicts how tightly each compound fits, in kcal/mol. Lower = tighter.
Rank hits, inspect poses in ChimeraX, compare against known binders.
Re-score with Boltz-2; send top hits to a mentor’s wet lab (MST / ITC).
All on a laptop. This is the CRS DCLK1 project.
Everything below is free with a school email — you don’t need to buy anything to start.
Thousands of published, downloadable datasets — each one a stack of unanswered questions. A few to start with:
Full, curated list lives on slcancerresearch.org.
Skim reviews until you can explain the area out loud.
Read 20+ papers, find the gap, make it FINER.
Send many; choose the technique you can actually run.
Run the analysis; keep a dated notebook of everything.
Make figures, then an OncoLegacy article, poster, or submission.
Months, not years. Consistency beats intensity.
Date everything. Reproducibility is the line between a hobby and research.
A small completed project beats a grand abandoned one. Scope down until it’s doable.
Reply fast, hit deadlines, never run code you can’t explain. Mentors remember this.
Publish in OncoLegacy, post updates. Visibility compounds — one project opens the next.
“Why?” and “What if?” beat “What is?” Curiosity is the whole job.
If you commit to a mentor, follow through. The research world is small.
OncoLegacy (our journal) · student research journals · preprints.
The CRS symposium · poster sessions · local science fairs.
ISEF · Regeneron STS · JSHS · regional fairs.
And AACR — the professional stage our Orlando team is aiming for. A finished project is your ticket.
A specific, answerable gap in the literature.
A tool you can run, and someone to guide it.
Data, a figure, a defensible finding.
OncoLegacy article, poster, abstract, AACR.
Do it ethically — IRB, honest data, real authorship. (More at our Ethics masterclass.)
Don't wait until you "know enough" — you learn research by doing it. Pick a topic tonight, and let CRS handle the hard part: finding you a question and a mentor.
Apply year-round · reviewed in 2–3 days · slcancerresearch.org
Then go find the answer nobody has yet.