Clonal Evolution

Mutational Signature

Properties1
tagsmutational-processes

4 min read

Mutational Signature

Definition

A mutational signature is a characteristic pattern of somatic mutations — defined by the relative frequencies of different substitution types in their trinucleotide sequence context — that reflects the activity of a specific DNA damage or repair process during a cell’s evolutionary history. Mutational signatures are extracted from the full catalog of somatic mutations in a cancer genome using mathematical decomposition methods such as non-negative matrix factorization (Nik-Zainal et al., 2012).

Rationale

The key insight is that passenger-mutations, which constitute the vast majority of somatic mutations in a cancer, preserve an unbiased record of the mutational processes that have been active throughout the tumor’s lifetime. As Nik-Zainal et al. (2012) explain: “These bystanders bear the imprints of the DNA damage and repair processes operative during the development of the cancer, unmodified by selection” (p. 980). Historically, mutation pattern analysis was restricted to cancer genes like TP53, where selective effects confound the mutational signal. Whole-genome sequencing of passengers bypasses this limitation.

Extraction Method

Non-negative matrix factorization (NMF) decomposes the matrix of mutation counts (96 substitution types × samples) into a set of mutational signatures, each defined by its characteristic profile across the 96 trinucleotide contexts, and a matrix of exposure estimates reflecting the activity of each signature in each sample (Nik-Zainal et al., 2012).

Key Signatures Identified

Nik-Zainal et al. (2012) demonstrated the approach in 21 breast cancers, identifying signatures including:

  • Signature 1A/B: C>T transitions at NpCpG dinucleotides — associated with age-related spontaneous deamination of 5-methylcytosine
  • Signature 3: Characteristic of BRCA1/2-deficient cancers — associated with defective homologous recombination repair
  • Signatures 2 and 13: C>T and C>G mutations at TpC dinucleotides — attributed to APOBEC-mutagenesis
  • The study also discovered kataegis — localized clusters of hypermutation colocalizing with rearrangement breakpoints

Temporal Dynamics

Gerstung et al. (2020) demonstrated that the mutational spectrum changes significantly throughout tumor evolution in approximately 40% of samples. Some signatures are more active early (clonal), others late (subclonal), implying changing exposures or repair deficiencies over the tumor’s lifetime. This finding means that a single tumor sample provides only a time-averaged view of the mutational processes that have shaped it.

The Flaw Is Source Code

The conceptual framework of productive-error (Extended Brain, 2026, adapting Buehler’s materials science) reframes mutational signatures in a specific way: each signature is a crack pattern in the genome — a structured diagnostic readout of the mutational process that caused it, encoding the identity and mechanism of the process in the specific pattern of the damage.

The mapping is not metaphorical but structural. When a material cracks, the crack pattern encodes the stress that produced it: where the crack initiated, which planes it followed, and how it terminated. When a genome accumulates mutations, the mutational pattern encodes the mutagenic process: which bases were targeted (substitution type), what sequence context was preferred (trinucleotide context), whether the mutagen acted on single-stranded or double-stranded DNA (strand bias, informing whether transcription-coupled repair was operative), and whether mutations were gradual or catastrophic (spatial clustering; see kataegis). Each dimension of the mutational signature is a dimension of the crack pattern — a dimension of diagnostic information.

This framing clarifies why passenger-mutations are informative precisely because they are unselected: they are the raw crack pattern, unmodified by the smoothing effect of selection on driver-mutations. Just as a materials scientist reads the crack pattern before polishing removes it, the cancer genomicist reads the mutational signature before selection distorts it. The signature IS the flaw-as-source-code.

The diagnostic value operates in both directions. A known mutagenic process predicts its signature (APOBEC → C>T and C>G at TpC). An observed signature identifies an unknown process (signatures of unknown etiology in the COSMIC catalog are crack patterns awaiting interpretation). The signature is the message.

Graph View

Backlinks