DEEPOMICS FFPE, a deep neural network model, identifies DNA sequencing artifacts from formalin fixed paraffin embedded tissue with high accuracy.

Formalin-fixed, paraffin-embedded (FFPE) tissue specimens are routinely used in pathological diagnosis, but their large number of artifactual mutations complicate the evaluation of companion diagnostics and analysis of next-generation sequencing data. Identification of variants with low allele frequencies is challenging because existing FFPE filtering tools label all low-frequency variants as artifacts. To address this problem, we aimed to develop DEEPOMICS FFPE, an AI model that can classify a true variant from an artifact. Paired whole exome sequencing data from fresh frozen and FFPE samples from 24 tumors were obtained from public sources and used as training and validation sets at a ratio of 7:3. A deep neural network model with three hidden layers was trained with input features using outputs of the MuTect2 caller. Contributing features were identified using the SHapley Additive exPlanations algorithm and optimized based on training results. The performance of the final model (DEEPOMICS FFPE) was compared with those of existing models (MuTect filter, FFPolish, and SOBDetector) by using well-defined test datasets. We found 41 discriminating properties for FFPE artifacts. Optimization of property quantification improved the model performance. DEEPOMICS FFPE removed 99.6% of artifacts while maintaining 87.1% of true variants, with an F1-score of 88.3 in the entire dataset not used for training, which is significantly higher than those of existing tools. Its performance was maintained even for low-allele-fraction variants with a specificity of 0.995, suggesting that it can be used to identify subclonal variants. Different from existing methods, DEEPOMICS FFPE identified most of the sequencing artifacts in the FFPE samples while retaining more of true variants, including those of low allele frequencies. The newly developed tool DEEPOMICS FFPE may be useful in designing capture panels for personalized circulating tumor DNA assay and identifying candidate neoepitopes for personalized vaccine design. DEEPOMICS FFPE is freely available on the web ( http://deepomics.co.kr/ffpe ) for research.

Breast Cancer Index and Prediction of Extended Aromatase Inhibitor Therapy Benefit in Hormone Receptor-Positive Breast Cancer from the NRG Oncology/NSABP B-42 Trial.

PURPOSE: BCI (H/I) has been shown to predict extended endocrine therapy (EET) benefit. We examined BCI (H/I) for EET benefit prediction in NSABP B-42, which evaluated extended letrozole therapy (ELT) in patients with hormone receptor-positive breast cancer after 5 years of ET. EXPERIMENTAL DESIGN: A stratified Cox model was used to analyze RFI as the primary endpoint, with DR, BCFI, and DFS as secondary endpoints. Because of a nonproportional effect of ELT on DR, time-dependent analyses were performed. RESULTS: The translational cohort included 2,178 patients (45% BCI (H/I)-High, 55% BCI (H/I)-Low). ELT showed an absolute 10-year RFI benefit of 1.6% (P = 0.10), resulting in an underpowered primary analysis (50% power). ELT benefit and BCI (H/I) did not show a significant interaction for RFI (BCI (H/I)-Low: 10 years absolute benefit 1.1% [HR, 0.70; 95% confidence interval (CI), 0.43-1.12; P = 0.13]; BCI (H/I)-High: 2.4% [HR, 0.83; 95% CI, 0.55-1.26; P = 0.38]; Pinteraction = 0.56). Time-dependent DR analysis showed that after 4 years, BCI (H/I)-High patients had significant ELT benefit (HR = 0.29; 95% CI, 0.12-0.69; P < 0.01), whereas BCI (H/I)-Low patients were less likely to benefit (HR, 0.68; 95% CI, 0.33-1.39; P = 0.29; Pinteraction = 0.14). Prediction of ELT benefit by BCI (H/I) was more apparent in the HER2- subset after 4 years (ELT-by-BCI (H/I) Pinteraction = 0.04). CONCLUSIONS: BCI (H/I)-High versus BCI (H/I)-Low did not show a statistically significant difference in ELT benefit for the primary endpoint (RFI). However, in time-dependent DR analysis, BCI (H/I)-High patients experienced statistically significant benefit from ELT after 4 years, whereas (H/I)-Low patients did not. Because BCI (H/I) has been validated as a predictive marker of EET benefit in other trials, additional follow-up may enable further characterization of BCI's predictive ability.

Phase III Randomized, Placebo-Controlled Trial of Endocrine Therapy ± 1 Year of Everolimus in Patients With High-Risk, Hormone Receptor-Positive, Early-Stage Breast Cancer.

PURPOSE: Phosphatidylinositol 3-kinase/AKT-serine threonine kinase/mammalian target of rapamycin (mTOR) pathway abnormalities contribute to endocrine resistance. Everolimus, an mTOR inhibitor, improved progression-free survival in hormone receptor-positive metastatic breast cancer (BC) when combined with endocrine therapy (ET). In this phase III randomized, placebo-controlled trial, we assessed the efficacy of everolimus + ET as adjuvant therapy in high-risk, hormone receptor-positive, human epidermal growth factor receptor 2-negative BC after adjuvant/neoadjuvant chemotherapy. METHODS: Patients were randomly assigned 1:1 to physician's choice ET and 1 year of everolimus (10 mg orally once daily) or placebo stratified by risk group. The primary end point was invasive disease-free survival (IDFS) evaluated by a stratified log-rank test with the hazard ratio (HR) estimated by Cox regression. Subset analyses included preplanned evaluation by risk group and exploratory analyses by menopausal status and age. Secondary end points included overall survival (OS) and safety. Everolimus did not improve IDFS/OS when added to ET in patients with early-stage high-risk, hormone receptor-positive BC. RESULTS: One thousand and nine hundred thirty-nine patients were randomly assigned with 1,792 eligible for analysis. Overall, no benefit of everolimus was seen for IDFS (HR, 0.94 [95% CI, 0.77 to 1.14]) or OS (HR, 0.97 [95% CI, 0.75 to 1.26]). The assumption of proportional hazards was not met suggesting significant variability in the HR over time since the start of treatment. In an unplanned subgroup analysis among postmenopausal patients (N = 1,221), no difference in IDFS (HR, 1.08 [95% CI, 0.86 to 1.36]) or OS (HR, 1.19 [95% CI, 0.89 to 1.60]) was seen. In premenopausal patients (N = 571), everolimus improved both IDFS (HR, 0.64 [95% CI, 0.44 to 0.94]) and OS (HR, 0.49 [95% CI, 0.28 to 0.86]). Treatment completion rates were lower in the everolimus arm compared with placebo (48% v 73%) with higher grade 3 and 4 adverse events (35% v 7%). CONCLUSION: One year of adjuvant everolimus + ET did not improve overall outcomes. Subset analysis suggests mTOR inhibition as a possible target for patients who remain premenopausal after chemotherapy.

Machine Learning Predicts Oxaliplatin Benefit in Early Colon Cancer.

PURPOSE: A combination of fluorouracil, leucovorin, and oxaliplatin (FOLFOX) is the standard for adjuvant therapy of resected early-stage colon cancer (CC). Oxaliplatin leads to lasting and disabling neurotoxicity. Reserving the regimen for patients who benefit from oxaliplatin would maximize efficacy and minimize unnecessary adverse side effects. METHODS: We trained a new machine learning model, referred to as the colon oxaliplatin signature (COLOXIS) model, for predicting response to oxaliplatin-containing regimens. We examined whether COLOXIS was predictive of oxaliplatin benefits in the CC adjuvant setting among 1,065 patients treated with 5-fluorouracil plus leucovorin (FULV; n = 421) or FULV + oxaliplatin (FOLFOX; n = 644) from NSABP C-07 and C-08 phase III trials. The COLOXIS model dichotomizes patients into COLOXIS+ (oxaliplatin responder) and COLOXIS- (nonresponder) groups. Eight-year recurrence-free survival was used to evaluate oxaliplatin benefits within each of the groups, and the predictive value of the COLOXIS model was assessed using the P value associated with the interaction term (int P) between the model prediction and the treatment effect. RESULTS: Among 1,065 patients, 526 were predicted as COLOXIS+ and 539 as COLOXIS-. The COLOXIS+ prediction was associated with prognosis for FULV-treated patients (hazard ratio [HR], 1.52 [95% CI, 1.07 to 2.15]; P = .017). The model was predictive of oxaliplatin benefits: COLOXIS+ patients benefited from oxaliplatin (HR, 0.65 [95% CI, 0.48 to 0.89]; P = .0065; int P = .03), but COLOXIS- patients did not (COLOXIS- HR, 1.08 [95% CI, 0.77 to 1.52]; P = .65). CONCLUSION: The COLOXIS model is predictive of oxaliplatin benefits in the CC adjuvant setting. The results provide evidence supporting a change in CC adjuvant therapy: reserve oxaliplatin only for COLOXIS+ patients, but further investigation is warranted.

Utility of the 70-Gene MammaPrint Assay for Prediction of Benefit From Extended Letrozole Therapy in the NRG Oncology/NSABP B-42 Trial.

PURPOSE: MammaPrint (MP) determines distant metastatic risk and may improve patient selection for extended endocrine therapy (EET). This study examined MP in predicting extended letrozole therapy (ELT) benefit in patients with early-stage breast cancer (BC) from the NSABP B-42 trial. PATIENTS AND METHODS: MP was tested in 1,866 patients randomly assigned to receive ELT or placebo. The primary end point was distant recurrence (DR). Secondary end points were disease-free survival (DFS) and BC-free interval (BCFI). Tumors were classified as MP high risk (MP-HR) or low risk (MP-LR). MP-LR tumors were further classified as ultralow risk (MP-UL) or low non-ultralow risk (MP-LNUL). RESULTS: There was no statistically significant difference in ELT benefit on DR between MP-HR and MP-LR (interaction P = .38). MP-LR tumors (n = 1,160) exhibited a statistically significant 10-year benefit of 3.7% for DR (hazard ratio [HR], 0.43 [95% CI, 0.25 to 0.74]; P = .002), whereas MP-HR tumors (n = 706) exhibited a nonsignificant 2.4% benefit (HR, 0.65 [95% CI, 0.34 to 1.24]; P = .19). The 10-year ELT benefit was significant for DFS (7.8%) and BCFI (7.0%) for MP-LR tumors, whereas MP-HR tumors did not significantly benefit (interaction DFS: P = .015, BCFI: P = .006). In exploratory analysis, the 10-year ELT benefit was significant and more pronounced in MP-LNUL (n = 908) tumors: 4.0% for DR, 9.5% for DFS, and 7.9% for BCFI; the benefit in MP-UL (n = 252) tumors was not significant: 3% for DR, 1.8% for DFS, and 4.1% for BCFI. CONCLUSION: The primary hypothesis of predictive ability of MP on DR was not confirmed. However, the secondary outcomes demonstrated MP was predictive of ELT response and identified a subset of patients with early-stage hormone receptor-positive BC (MP-LR) with improved outcomes from ELT. These data could have important clinical implications in patient selection beyond clinical risk assessment for EET.