Thirty-Three Uterine Cancers, Sequenced: A Genomic Atlas Reveals the Molecular Fault Lines of Endometrial Disease

Deep sequencing of 33 endometrial tumors at Memorial Sloan Kettering maps the mutations, microsatellite instability patterns, and survival trajectories that define one of the most common gynecologic cancers.

Every year, roughly 66,000 women in the United States are diagnosed with cancer of the uterine lining — endometrial cancer. For many, the prognosis is good: early-stage endometrioid tumors, the most common subtype, are among the most curable solid cancers. But the disease is not one thing. Lurking within the same organ are serous carcinomas that behave more like aggressive ovarian cancers, clear cell variants with their own biology, and a subset of tumors so riddled with mutations that their DNA repair machinery has essentially collapsed. Telling these molecular subtypes apart matters enormously, because the treatments that work for one can fail entirely for another.

This dashboard dissects the genomic landscape of 33 female patients whose endometrial tumors were profiled on Memorial Sloan Kettering's MSK-IMPACT targeted sequencing panels. Each tumor's somatic mutations, copy-number changes, microsatellite instability score, tumor mutational burden, and overall survival are laid out in interactive charts — a compact but revealing portrait of a disease defined by its diversity.

Who Is in the Cohort

Before any genomic finding can be trusted, the data underneath it must be clean. The cohort comprises exactly 33 patients, each represented once in the clinical files — a fact confirmed by a simple identity check across the patient roster. All samples originate from the uterus, all patients are female, and the sequencing was performed on one of two versions of the MSK-IMPACT panel: IMPACT410 (covering 410 cancer-related genes) or IMPACT468 (an expanded version with roughly 60 additional genes). That panel split matters. Any gene present only on IMPACT468 will appear wild-type in every IMPACT410 sample — not because it lacks mutations, but because it was never interrogated. Downstream mutation-frequency comparisons must be restricted to the 410-gene intersection to avoid false negatives.

Histologically, the cohort includes endometrioid, clear cell, and other subtypes, with OncoTree diagnostic codes providing finer-grained classification. Endometrioid cases dominate, as expected for any unselected endometrial cancer series, but the presence of rarer histologies adds molecular contrast.

The Two Worlds of Mutational Burden

Tumor mutational burden — the number of nonsynonymous mutations per megabase of sequenced DNA — is not evenly distributed across these 33 tumors. The distribution is strikingly bimodal. A cluster of samples sits at low TMB, consistent with microsatellite-stable (MSS) tumors whose genomes are driven more by copy-number changes than point mutations. Then there is a long right tail: hypermutated tumors carrying far more mutations than their stable counterparts.

This split is not random. It maps directly onto microsatellite instability. MSI-H tumors — those whose mismatch repair system has failed, either through Lynch syndrome germline mutations or sporadic silencing of the MLH1 gene — accumulate mutations at microsatellite repeat sequences throughout the genome. The continuous MSI sensor score, plotted alongside TMB, shows the two populations separating cleanly: low-MSI/low-TMB in one corner, high-MSI/high-TMB in the other. Samples exceeding 10 mutations per megabase cross the FDA-approved threshold for pembrolizumab eligibility — a clinically actionable line.

The distribution is strikingly bimodal — a cluster of stable tumors at low TMB, and a hypermutated tail where DNA repair has essentially collapsed.

The Driver Landscape: PTEN, PIK3CA, and the Usual Suspects

Endometrial cancer has a well-established cast of driver genes, and this cohort recapitulates it faithfully. The oncoprint — a gene-by-sample matrix showing the top 25 recurrently altered genes — places PTEN, PIK3CA, TP53, ARID1A, and CTNNB1 near the top, each altered in a substantial fraction of patients.

The patterns within that matrix tell a deeper story. PTEN loss and PIK3CA mutation frequently co-occur in the same tumors, reflecting their shared role in the PI3K/AKT signaling pathway — a convergence that makes the pathway an attractive therapeutic target. TP53 mutations, by contrast, tend to appear in samples that lack PTEN alterations, consistent with the known mutual exclusivity between the endometrioid (PTEN-driven) and serous (TP53-driven) molecular subtypes.

The variant classification spectrum reinforces the MSI-H signature: an elevated proportion of frameshift insertions and deletions, the hallmark of mismatch repair deficiency, sits alongside the expected missense-dominant background. The SNP-to-indel ratio is shifted toward indels more than in most solid tumors — a direct consequence of slippage at microsatellite repeats.

Survival and Its Limits at Small Scale

Overall survival data are available for all 33 patients, with death events and censoring times recorded. The swimmer plot — horizontal bars stretching from enrollment to last follow-up or death — reveals a mix of outcomes: some patients with long bars and no event (alive and censored), others with short bars ending in a death marker. The Kaplan-Meier curve for the full cohort provides a median survival estimate, though the small sample size means that the tail of the curve, where only a handful of patients remain at risk, becomes unreliable.

Stratifications by self-reported race and ethnicity probe whether the well-documented disparities in endometrial cancer outcomes — Black women experience measurably worse survival nationally — are detectable in this cohort. With only 33 patients, racial subgroups are small enough that any apparent separation between curves must be treated as hypothesis-generating rather than conclusive. The sex-stratified survival curve, meanwhile, serves as a metadata integrity check: because endometrial cancer occurs almost exclusively in individuals with a uterus, a single female curve should appear. A second curve would flag a data annotation error.

Quality Control: Purity and Read Depth

Every somatic mutation call in the dashboard depends on sequencing quality, and two metrics matter most: tumor purity and read depth. Tumor purity — the fraction of DNA in a specimen that comes from cancer cells rather than admixed normal tissue — determines whether a mutation present in the tumor will be detectable. Samples below 20% purity risk missing even clonal mutations, because the variant allele fraction drops below the caller's sensitivity threshold. The purity distribution across the cohort flags any such specimens.

The reference-versus-alternate read count plot offers a complementary view. Each somatic variant is a dot: its position along the x-axis reflects how many reads matched the normal reference allele, and along the y-axis, how many carried the mutant allele. Variants clustering at high depth on both axes confirm adequate MSK-IMPACT coverage. Points hugging the x-axis — high reference, very low alternate — represent calls at the detection limit, where confidence is lowest. Any sample appearing mostly wild-type on the oncoprint but showing low purity on this tab may simply lack the sensitivity to detect its mutations, rather than truly lacking them.

Thirty-three patients is not a large cohort. It is too small to power definitive survival comparisons by race, too small to detect rare driver genes, too small to resolve every molecular subtype with confidence. But it is large enough to do something important: to demonstrate that even a compact, well-characterized dataset can recapitulate the known molecular architecture of endometrial cancer — the PTEN-PIK3CA axis, the MSI-H hypermutation signature, the TP53-serous divide — and do so at single-patient resolution. Each of these 33 genomes is a clinical decision point for a real patient. The open question is whether the patterns visible here, particularly the discordant outliers with high TMB but low MSI scores, point toward molecular subtypes — POLE-ultramutated tumors, for instance — that targeted panels can now identify prospectively. As MSK-IMPACT data accumulate across thousands of endometrial cases, cohorts like this one become the building blocks of a larger atlas, one where sample size catches up to biological complexity.