Cancer: When Control Fails

Cancer is not one disease. It's a general process that can occur in virtually any tissue: the accumulation of somatic mutations that disable normal growth controls, enabling cells to proliferate without restraint, invade surrounding tissue, and eventually colonize distant organs. The underlying mechanism — Darwinian evolution operating on somatic cells — is universal. The specific mutations, the affected pathways, and the clinical behavior vary enormously between cancer types.

Understanding cancer biology computationally means understanding selection pressures on tumors, clonal evolution, the landscape of driver mutations, and why cancers are so hard to cure. These concepts directly inform how you design cancer genomics analyses, interpret results, and think about therapeutic strategies.

The Clonal Evolution Model

Cancer begins with a single cell. One cell acquires a mutation that provides a growth advantage over its neighbors — faster proliferation, resistance to apoptosis, or escape from growth factor dependence. That cell divides, producing daughter cells that inherit the mutation. Subsequent mutations accumulate in the growing population, and cells that acquire additional growth-promoting mutations outcompete their peers.

This is Darwinian evolution operating inside the body, with somatic cells as the unit of selection:

• Heritable variation: somatic mutations generate variants
• Selection: cells with growth advantages proliferate faster
• Differential reproduction: faster-dividing clones expand at the expense of slower ones

Driver vs. Passenger Mutations

A tumor genome may contain thousands of somatic mutations. Most are passenger mutations — neutral changes that accumulated in the lineage of the cancer cell but don't contribute to the cancer phenotype. They're just along for the ride.

A small subset are driver mutations — variants that provide growth advantages and were positively selected during tumor evolution. Identifying drivers is the central challenge of cancer genomics.

How to distinguish drivers from passengers:

• Recurrence: a mutation found in many independent tumors of the same type is likely selected, not random
• Functional impact: affects a domain known to regulate cellular function (kinase domain, DNA-binding domain, tumor suppressor)
• Position: "hotspot" mutations at specific codons (e.g., KRAS G12, BRAF V600, TP53 R175/248/249/273) are strongly selected
• Comparative frequency: excess mutation rate at certain positions relative to background mutation rate

MutSigCV and dndscv are statistical tools that identify genes under positive selection in cancer by comparing the observed mutation rate with the expected background rate.

Oncogenes vs. Tumor Suppressors

Driver mutations fall into two functional categories with opposite effects:

Oncogenes

Oncogenes are gain-of-function mutations in genes that promote cell growth. The normal form is called a proto-oncogene (essential genes regulating growth); the mutated, constitutively active form is the oncogene.

Activation mechanisms:

• Point mutation: KRAS G12D locks KRAS in the GTP-bound active state
• Gene amplification: ERBB2 (HER2) amplification in breast cancer → overexpressed growth factor receptor
• Chromosomal translocation: BCR-ABL fusion in CML creates a constitutively active tyrosine kinase
• Promoter mutation: TERT promoter mutations in melanoma create new TF binding sites → increase telomerase expression

Dominant: only ONE allele needs to be mutated to activate the gene. The mutant allele produces a constitutively active protein even when the normal allele is present.

Tumor Suppressors

Tumor suppressor genes are loss-of-function mutations in genes that normally restrict cell growth. When both copies are inactivated, the growth brake is released.

Knudson's two-hit hypothesis (1971): tumor suppressors require inactivation of BOTH alleles. The first "hit" can be inherited or somatic; the second is somatic. In familial retinoblastoma, the first hit is inherited (germline), so only one somatic hit is needed in each cell → earlier, bilateral tumors. This model was later validated molecularly.

Inactivation mechanisms:

• Point mutation (nonsense, frameshift, splice site): destroys protein function
• Deletion: removes the gene copy
• Promoter methylation (epigenetic silencing): turns off transcription
• Loss of heterozygosity (LOH): mitotic recombination or deletion removes the remaining wild-type allele

Major tumor suppressors:

• TP53 (~50% of all cancers): transcription factor; triggers arrest or apoptosis in response to damage
• RB1 (many cancers): cell cycle regulator; mutated → bypass of G1/S checkpoint
• PTEN (~30% of cancers): phosphatase; inactivation → constitutive PI3K/AKT signaling
• APC (colorectal cancer): negative regulator of Wnt/β-catenin signaling
• BRCA1/2: DNA repair by homologous recombination; germline mutations → hereditary breast/ovarian cancer

The Hallmarks of Cancer

Hanahan and Weinberg's Hallmarks of Cancer (2000, updated 2011 and 2022) provides a framework for thinking about the capabilities a cell must acquire to become malignant:

1. Sustained proliferative signaling — oncogenic KRAS, EGFR mutations; autocrine growth factor loops
2. Evasion of growth suppressors — RB pathway inactivation; contact inhibition loss
3. Resisting cell death — BCL-2 overexpression; p53 inactivation; anti-apoptotic survival signals
4. Enabling replicative immortality — telomerase reactivation; bypassing replicative senescence
5. Induction of angiogenesis — VEGF overexpression; hypoxia-driven neo-vascularization
6. Activating invasion and metastasis — E-cadherin loss; MMP expression; epithelial-mesenchymal transition
7. Reprogramming energy metabolism — Warburg effect (aerobic glycolysis)
8. Evading immune destruction — PD-L1 upregulation; immunosuppressive microenvironment
9. Unlocking phenotypic plasticity — dedifferentiation; stem cell-like states (2022 addition)
10. Nonmutational epigenetic reprogramming (2022 addition)
11. Polymorphic microbiomes (2022 addition)
12. Senescent cells (2022 addition)

No single tumor has mutations in all these hallmarks, but every successful cancer has addressed enough of them to proliferate, survive, and spread.

Tumor Heterogeneity and Clonal Evolution

A key insight from deep sequencing of tumors: they are not genetically homogeneous. A primary tumor may contain dozens of distinct subclones, each with its own mutation profile.

Tumor heterogeneity takes two forms:

• Spatial heterogeneity: different regions of the same tumor have different mutations (some clones are more aggressive; some are more drug-resistant)
• Temporal heterogeneity: the tumor evolves over time — treatment kills sensitive cells, resistant clones expand

This has critical implications for cancer treatment:

• A biopsy from one region may not represent the whole tumor (sampling bias)
• A drug that eliminates the dominant clone may allow a resistant subclone to expand
• Liquid biopsy (circulating tumor DNA from blood) can capture tumor heterogeneity more comprehensively than tissue biopsy

The cancer evolution perspective, pioneered by Mel Greaves and Charles Swanton, views cancer treatment as evolutionary medicine — designed to prevent resistance rather than just kill cells.

The Cancer Genome Landscape

Large-scale cancer genome sequencing projects (TCGA, ICGC, PCAWG) have revealed consistent patterns:

• Mutation burden varies enormously: from <1 somatic mutation/Mb (pediatric cancers) to >100 mutations/Mb (microsatellite-instable colorectal, UV-damaged melanoma)
• A small set of genes are recurrently mutated across cancer types: TP53, KRAS, PIK3CA, PTEN, RB1, APC, CDH1, ARID1A, FBXW7
• Most driver mutations affect a limited set of signaling pathways: cell cycle regulation (CDK4/6-RB), RAS/MAPK, PI3K/AKT/mTOR, p53/apoptosis, Wnt/β-catenin, TGF-β, receptor tyrosine kinases, chromatin regulation
• Mutational signatures reveal cancer etiology: UV in skin cancer, APOBEC in many cancers, tobacco in lung cancer, alcohol in head/neck/liver cancer

Synthetic Lethality: Exploiting Cancer's Weaknesses

One of the most important translational concepts: synthetic lethality occurs when the combination of two genetic defects is lethal, whereas either defect alone is tolerated.

In cancer, if the tumor has lost gene A (e.g., BRCA1), and gene B is required for survival only when gene A is absent, then inhibiting gene B will kill tumor cells but not normal cells (which have intact BRCA1).

PARP inhibitors exploit this in BRCA1/2-mutant cancers. BRCA1/2 are required for homologous recombination repair of double-strand DNA breaks. When BRCA1/2 are absent, cells rely on PARP-mediated base excision repair as a backup. PARP inhibitors (olaparib, niraparib, rucaparib) block this backup pathway → DNA damage accumulates → BRCA-mutant cancer cells die. Normal cells with intact BRCA1/2 are unaffected.

The computational approach to finding synthetic lethalities: CRISPR screens in cancer cell lines (which genes are essential in cancer cells but not in normal cells?), combined with genomic analysis of which tumor suppressors are lost in which cancers.

Why Cancers Are Hard to Cure

Three fundamental reasons:

1. Heterogeneity: the dominant clone killed by therapy is not the entire tumor. Subclones with different mutations survive, proliferate, and become the new dominant clone (acquired resistance).

2. Evolution: cancer cells mutate rapidly (especially with replication stress and MMR deficiency). Any drug creates a selection pressure that favors resistant variants that preexist in the population.

3. Resemblance to normal cells: cancer arises from normal cells by point mutations. The targets (growth signaling proteins, cell cycle regulators) often have essential functions in normal cells — limiting how aggressively they can be inhibited without toxicity.

Immunotherapy partially circumvents the third problem by leveraging the immune system's ability to discriminate (recognizing neoantigens that don't exist in normal cells), but tumors find ways to evade immunity too.

The goal of cancer genomics is to understand tumor biology well enough to outsmart the evolutionary pressure — to identify combinations of vulnerabilities that the tumor cannot simultaneously evade, given the constraints of its mutation landscape.