Performance of Lynch syndrome predictive models in quantifying the likelihood of germline mutations in patients with abnormal MLH1 immunoexpression

Genomics and integrative clinical data machine learning scoring model to ascertain likely Lynch syndrome patients

BACKGROUND

Lynch syndrome (LS) screening methods include multistep molecular somatic tumor testing to distinguish likely-LS patients from sporadic cases, which can be costly and complex. Also, direct germline testing for LS for every diagnosed solid cancer patient is a challenge in resource limited settings. We developed a unique machine learning scoring model to ascertain likely-LS cases from a cohort of colorectal cancer (CRC) patients.

METHODS

We used CRC patients from the cBioPortal database (TCGA studies) with complete clinicopathologic and somatic genomics data. We determined the rate of pathogenic/likely pathogenic variants in five (5) LS genes (MLH1, MSH2, MSH6, PMS2, EPCAM), and the BRAF mutations using a pre-designed bioinformatic annotation pipeline. Annovar, Intervar, Variant Effect Predictor (VEP), and OncoKB software tools were used to functionally annotate and interpret somatic variants detected. The OncoKB precision oncology knowledge base was used to provide information on the effects of the identified variants. We scored the clinicopathologic and somatic genomics data automatically using a machine learning model to discriminate between likely-LS and sporadic CRC cases. The training and testing datasets comprised of 80% and 20% of the total CRC patients, respectively. Group regularisation methods in combination with 10-fold cross-validation were performed for feature selection on the training data.

RESULTS

Out of 4800 CRC patients frorm the TCGA datasets with clinicopathological and somatic genomics data, we ascertained 524 patients with complete data. The scoring model using both clinicopathological and genetic characteristics for likely-LS showed a sensitivity and specificity of 100%, and both had the maximum accuracy, area under the curve (AUC) and AUC for precision-recall (AUCPR) of 1. In a similar analysis, the training and testing models that only relied on clinical or pathological characteristics had a sensitivity of 0.88 and 0.50, specificity of 0.55 and 0.51, accuracy of 0.58 and 0.51, AUC of 0.74 and 0.61, and AUCPR of 0.21 and 0.19, respectively.

CONCLUSIONS

Simultaneous scoring of LS clinicopathological and somatic genomics data can improve prediction and ascertainment for likely-LS from all CRC cases. This approach can increase accuracy while reducing the reliance on expensive direct germline testing for all CRC patients, making LS screening more accessible and cost-effective, especially in resource-limited settings.

Performance evaluation of predictive models for detecting MMR gene mutations associated with Lynch syndrome in cancer patients in a Chinese cohort in Taiwan

Identifying Lynch syndrome significantly impacts cancer risk management, treatment, and prognosis. Validation of mutation risk predictive models for mismatch repair (MMR) genes is crucial for guiding genetic counseling and testing, particularly in the understudied Asian population. We evaluated the performance of four MMR mutation risk predictive models in a Chinese cohort of 604 patients with colorectal cancer (CRC), endometrial cancer (EC), or ovarian cancer (OC) in Taiwan. All patients underwent germline genetic testing and 36 (6.0%) carried a mutation in the MMR genes (MLH1, MSH2, MSH6, and PMS2). All models demonstrated good performance, with area under the receiver operating characteristic curves comparable to Western cohorts: PREMM5 0.80 (95% confidence interval [CI], 0.73-0.88), MMRPro 0.88 (95% CI, 0.82-0.94), MMRPredict 0.82 (95% CI, 0.74-0.90), and Myriad 0.76 (95% CI, 0.67-0.84). Notably, MMRPro exhibited exceptional performance across all subgroups regardless of family history (FH+ 0.88, FH- 0.83), cancer type (CRC 0.84, EC 0.85, OC 1.00), or sex (male 0.83, female 0.90). PREMM5 and MMRPredict had good accuracy in the FH+ subgroup (0.85 and 0.82, respectively) and in CRC patients (0.76 and 0.82, respectively). Using the ratio of observed and predicted mutation rates, MMRPro and PREMM5 had good overall fit, while MMRPredict and Myriad overestimated mutation rates. Risk threshold settings in different models led to different positive predictive values. We suggest a lower threshold (5%) for recommending genetic testing when using MMRPro, and a higher threshold (20%) when using PREMM5 and MMRPredict. Our findings have important implications for personalized mutation risk assessment and counseling on genetic testing.

PREMM5 distinguishes sporadic from Lynch syndrome-associated MMR-deficient/MSI-high colorectal cancer

Current algorithms for diagnosing Lynch syndrome (LS) include multistep molecular tumor tests to distinguish LS-associated from sporadic colorectal cancer (CRC), which add cost and complexity to the evaluation. We hypothesized that PREMM5, a clinical LS prediction tool, could be an alternative approach to screen for LS, thereby lessening the need for specialized molecular diagnostics. We reviewed a consecutively ascertained institutional cohort of 1058 CRC patients on whom pathologic and clinical data were available, including prior LS germline testing. Data from MMR-D/MSI-H CRC patients were reviewed and PREMM5 scores were calculated for each individual. Using a PREMM5 score cutoff ≥ 2.5% to characterize the need for germline testing, we determined the rate of pathogenic/likely pathogenic germline variants (PGVs) in LS genes in patients with PREMM5 scores ≥ 2.5% versus < 2.5%. Sensitivity and negative predictive values (NPV) of PREMM5 were calculated for all MMR-D/MSI-H CRC patients, and those with MLH1-deficient CRC. MMR IHC and/or MSI results were available on 572/1058 cases. We identified 74/572 (12.9%) cases as MMR-D/MSI-H, of which 28/74 (37.8%) harbored a LS PGV. 11/49 (22.4%) patients with MLH1-deficient CRC harbored a LS PGV. PREMM5 had 100% sensitivity (95% CI: 87.7-100 for any MMR-D/MSI-H; 95% CI: 71.5-100 for MLH1-deficient CRC) and 100% NPV (95% CI: 83.2-100 for any MMR-D/MSI-H; 95% CI: 82.4-100 for MLH1-deficient CRC) for identifying LS PGVs in these cohorts. PREMM5 accurately distinguishes LS- from non-LS-associated MMR-D/MSI-H CRC without additional somatic molecular testing. These findings are particularly relevant for limited-resource settings where advanced molecular diagnostics may be unavailable.

Contribution of MLH1 constitutional methylation for Lynch syndrome diagnosis in patients with tumor MLH1 downregulation

Constitutional epimutation of the two major mismatch repair genes, MLH1 and MSH2, has been identified as an alternative mechanism that predisposes to the development of Lynch syndrome. In the present work, we aimed to investigate the prevalence of MLH1 constitutional methylation in colorectal cancer (CRC) patients with abnormal expression of the MLH1 protein in their tumors. In a series of 38 patients who met clinical criteria for Lynch syndrome genetic testing, with loss of MLH1 expression in the tumor and with no germline mutations in the MLH1 gene (35/38) or with tumors presenting the BRAF p.Val600Glu mutation (3/38), we screened for constitutional methylation of the MLH1 gene promoter using methylation‐specific multiplex ligation‐dependent probe amplification (MS‐MLPA) in various biological samples. We found four (4/38; 10.5%) patients with constitutional methylation in the MLH1 gene promoter. RNA studies demonstrated decreased MLH1 expression in the cases with constitutional methylation when compared with controls. We could infer the mosaic nature of MLH1 constitutional hypermethylation in tissues originated from different embryonic germ layers, and in one family we could show that it occurred de novo. We conclude that constitutional MLH1 methylation occurs in a significant proportion of patients who have loss of MLH1 protein expression in their tumors and no MLH1 pathogenic germline mutation. Furthermore, we provide evidence that MLH1 constitutional hypermethylation is the molecular mechanism behind about 3% of Lynch syndrome families diagnosed in our institution, especially in patients with early onset or multiple primary tumors without significant family history.