21P Scalable translational research: A novel extraction free whole transcriptome analyses approach specifically for low quality FFPE diagnostic samples which failed QC for standard NGS analyses in a TNBC cohort (VISION study)

Background: Invasive lobular carcinoma (ILC) and invasive ductal carcinoma (IDC) differ in biology and clinical behavior, yet mixed or ambiguous histologies and interobserver variability may obscure underlying tumor biology.The clinical relevance of histology-transcriptome discordance remains incompletely defined. Methods:We developed a sparse logistic-regression-based transcriptomic ILClikeness score in TCGA and validated it in METABRIC and two independent GEO cohorts.Beyond discrimination of pure subtypes, the score was applied to ambiguous tumors and to identify histology-transcriptome discordance.Discordant groups were biologically characterized by differences in CDH1 mutation frequency and expression.Prognostic significance was evaluated using Kaplan-Meier and multivariable Cox regression adjusted for key clinical covariates. Results:The score robustly discriminated ILC from IDC in METABRIC (AUC 0.86) and 3 independent GEO cohorts (AUC 0.86-0.92)with a total of 2,654 patients.Mixed or ambiguous histologies showed intermediate ILC-likeness scores between pure IDC and ILC.Histologically ductal tumors with ILC-like transcriptomic profiles were biologically distinct from concordant IDC, demonstrating reduced CDH1 expression (p<0.001) and enriched CDH1 mutations (6.3% vs 0.9%; OR 7.55, p=0.007).Among histologically defined IDC, patients with ILC-like transcriptomic profiles had significantly different overall survival compared with concordant IDC (log-rank p=0.001), which remained significant in multivariable analysis. Conclusions:The ILC-likeness signature identifies lobular-like biology within histologically ductal breast cancers, supported by CDH1 alterations and survival differences.These findings may inform risk stratification beyond histology.