Cell-free orphan noncoding RNAs and AI enable early detection of invasive breast cancer and ductal carcinoma in situ

Background

• Earlier detection of breast cancer through mammography screening has reduced disease-specific mortality.
• Confounding issues such as breast density, tumor size, and imaging anomalies can result in false negatives and ultimately later stage diagnosis.
• We have previously demonstrated high sensitivity and specificity for early detection of invasive breast cancer (BC) by utilizing a novel category of cancer-associated small RNAs, termed orphan noncoding RNAs (oncRNAs), through a liquid biopsy platform.
• In this study, we further assess the diagnostic ability of oncRNAs in a larger cohort of invasive BC and ductal carcinoma in situ (DCIS) for early cancer detection.

Goals

• Further improve our ability to detect invasive BC in a larger, multi-source cohort and demonstrate that the same platform has the potential to also detect DCIS.
• Establish that this next generation liquid biopsy platform has the ability to complement mammography and enable earlier detection of breast cancer.

Samples

• The study cohort includes clinically diagnosed female breast cancer patients (N=279) and age- and sex-matched individuals with no known diagnosis of cancer (N=304).
• Samples were sourced from Indivumed (Hamburg, Germany), Proteogenex (Inglewood, CA), and MT Group (Los Angeles, CA).
• Breast cancer samples were acquired at time of diagnosis and were treatment-naive. All cancer and cancer-free samples were collected between 2010-2022.
• Patients provided informed consent and contributing centers obtained IRB approval.

Methods

• The small RNA content of all samples was sequenced at an average depth of 25.28 ± 9.37 million 50-bp single-end reads.
• These reads were annotated using a library of 20,538 oncRNAs discovered through The Cancer Genome Atlas (TCGA) small RNA-seq database. These oncRNAs were found to be significantly enriched among 1,103 breast tumors compared to 349 normal tissue samples spanning multiple tissue sites through a female-specific analysis.
• We detected 18,025 (87.8%) unique breast cancer-specific oncRNA species within at least one sample from the study cohort.
• We then trained a generative AI model using 5-fold cross-validation to predict cancer status for all samples.

Overview of oncRNA Profiling and AI-Driven Model for Breast Cancer Prediction

Figure 1. Schematic of oncRNA Profiling and Modeling Pipeline

Our generative AI model utilizes tumor-derived oncRNAs discovered in TCGA female tissue samples for downstream applications in serum.

Study Demographics

Ability of oncRNA-Based Model for Prediction of Overall Breast Cancer Status in Serum

Figure 2. Overall Model Performance by ROC and Sensitivity at 90% Specificity

• The ROC curve (A) demonstrated an AUC of 0.95 (95% CI: 0.94–0.97) in predicting the cancer status of both invasive BC and DCIS samples at a sensitivity of 0.85 (0.81–0.89).
• Sensitivities at 90% specificity can be seen by cancer stage (B) and tumor T-category (C). 95% confidence intervals were calculated using the Clopper-Pearson method within each group.

Figure 3. Model ROC and Sensitivity in (A) DCIS and (B) Invasive BC

• Prediction of cancer within individuals with invasive BC compared to cancer-free controls resulted in an AUC of 0.95 (95% CI, 0.93–0.97) with a sensitivity of 0.87 (0.82–0.91) at 90% specificity.
• Within the DCIS cohort, we had an AUC of 0.93 (95% CI, 0.93– 0.97) with a sensitivity of 0.77 (0.82–0.91) at 90% specificity.

Conclusions