Two-Letter DNA Gaps Point to Platinum Treatment and Faulty Repair

Platinum drugs such as cisplatin, carboplatin, and oxaliplatin are a backbone of cancer treatment. They work by attaching to DNA and stalling the cell's copying machinery. A study in Nucleic Acids Research, published on October 13, 2026, asks what scars survive when tumor cells push past that damage. ^[1]

The finding

Missing DNA pairs in mutation catalogs

When a cell's DNA is damaged or miscopied, the error can stay behind as a somatic mutation, a change that arises in a tumor and not in inherited DNA. Different causes leave different patterns. Researchers read these patterns as mutational signatures: a fingerprint of which kinds of changes, in which surrounding letters, a process tends to produce. A large catalog from whole-genome sequencing includes signatures for pairs of adjacent letters changed together, called double-base substitutions, and for small insertions and deletions. ^[2]

Platinum drugs mostly bind two neighboring guanines (G, one of DNA's four letters). Yet the catalog's platinum-linked double-base signature, labeled DBS5, lacks the matching two-G pattern. ^[1] Earlier work on metastatic tumors had already tied platinum drugs to mutational footprints. ^[3] The authors suspected a missing piece: cells might skip a damaged pair entirely, leaving a deletion, and existing signatures do not record which two letters vanished. ^[1]

Reading additional clues from tumor genomes

Cancer genomes keep a record of what damaged them and how cells coped. Here, adding one overlooked event type to a standard counting scheme surfaced a treatment clue and a possible marker from data that already existed. Whether either holds up in new patients and in lab tests is the next question.

The evidence

Deletion signatures and treatment histories

The data. The team used whole-genome sequencing from 5,822 biopsies of metastatic tumors in the CPCT-02 study. Of these, 3,763 had a recorded treatment history, and the comparison selected samples from the tumor types in which platinum treatment was common: breast, esophagus, and ovary for carboplatin; lung and urothelial (urinary tract) for cisplatin; and colorectal for oxaliplatin. Each group was paired with untreated samples of the same tumor type. ^[1]

The method. They extended the standard list of 78 double-base substitutions with one extra option per letter pair, a deletion of the pair, giving 88 categories. They counted these per tumor and fed the counts to SigProfilerExtractor, a program that uses non-negative matrix factorization (NMF) to split many tumors' mutation counts into a few recurring signatures. ^{[1, 4]} Running treated and untreated groups separately lets a signature seen only after treatment stand out. Tumors differ widely by type, so pooling all cancers would blur signatures; the authors say they limited each analysis to selected tumor types for this reason. ^[1]

An oxaliplatin pattern. In colorectal tumors from oxaliplatin-treated patients, one signature (DBSD-2) combined deletion of a CC pair, the same event as GG on the opposite strand, with a known platinum-linked substitution. The authors report this is the first time a GG-type event appeared in a platinum-associated signature. This signature made up at least 10 percent of the double-base events in 56 percent of oxaliplatin-treated samples (152 of 270), against 33 percent (33 of 99) for carboplatin-treated and 21 percent (27 of 131) for cisplatin-treated samples. Its share was far lower in 1,063 mostly untreated primary colorectal cancers (p = 2.1e-24). ^[1]

Counting deletions without signatures told the same story. Two-base deletions were more common only in oxaliplatin-treated colorectal patients (odds ratio 23.2, where 1 means no difference). Among deletions of exactly two Gs, 95.1 percent (adjusted for sample size) came from oxaliplatin-treated colorectal samples, versus 67.0 and 62.1 percent in the carboplatin and cisplatin comparisons. Most of these deletions were two Gs standing alone, not part of a long run of Gs. The authors found no sign that transcription or replication direction explained them. ^[1]

A bonus: a marker for microsatellite instability. In every analysis, one signature was simply deletion of two Ts (TT). Microsatellite instability (MSI) means the cell's mismatch repair system, which fixes copying slips in repetitive DNA, is faulty. The TT signature tracked MSI in colorectal samples. In 1,063 primary colorectal cancers, a threshold of 1,000 TT deletions missed 2 of 223 MSI tumors and wrongly flagged 6 of 840 non-MSI tumors. Those MSI labels were inferred from genome features rather than a clinical test. In 24 colorectal cell lines, TT-deletion counts separated MSI from non-MSI. ^[1] Almost all TT deletions (98.3 percent) sat inside runs of more than six Ts. The authors also propose that an indel signature of unknown cause, ID12, reflects mismatch repair failure. ^[1]

A tiny blood test pilot. The team sequenced cell-free DNA (cfDNA), fragments of DNA floating in blood plasma, at about 5-fold coverage from three patients whose primary tumors had MSI, plus 41 healthy donors. One patient with a clear tumor signal in a droplet PCR test showed 6,846 TT deletions; two others showed 134 and 110, against a healthy average of 97.5. The authors call this a pilot and say a much larger cohort is needed before drawing thresholds. ^[1]

Why this approach. One reasonable view is that NMF fits because nobody knew how many platinum-specific patterns to expect, and matched untreated tumors supply the comparison. Lab experiments that dose cells and sequence them are a reasonable alternative that can test cause, but they were not part of this study.

The authors note that patients usually received platinum together with other drugs, so signatures could not be assigned to platinum alone. ^[1] The data are observational, so the links are associations. This post did not independently rerun any analysis.

Future directions

The authors call for a much larger cfDNA cohort to test the TT-deletion signal. An independent evaluation could compare those counts with established MSI labels, asking whether the association seen in the small blood pilot persists. The pilot does not establish a clinical test. ^[1]

The treated-versus-untreated tumor comparison also leaves causation unresolved. To distinguish a treatment association from a mutational mechanism, controlled exposure experiments could test whether platinum treatment produces the deletion pattern, using the paper's extended mutation categories to compare the resulting signatures. ^[1]