Research Group
Serum samples were collected from untreated BC patients employed at Shandong Cancer Hospital, a member of Shandong Daiichi Medical University and Shandong Academy of Sciences (March 2023 to March 2024). Controls were selected from physical examination departments that met the following criteria: There is no history of malignant tumors, benign breast neoplasms, or abnormal inflammation/physiological indicators (heart rate, blood pressure, temperature). Controls were age- and gender-matched in BC patients. All BC diagnoses followed the AJCC 8th Edition TNM staging criteria. CA153 levels were measured using a Roche Cobas E801 ImmunoAssay Analyzer (Roche, Shanghai, China). This study was approved by the ethics committee of Shandong Cancer Hospital, affiliated with Shandong Daiichi Medical University and Shandong Academy of Medicine.
Serum preparation and RNA extraction
Blood samples were collected in serum separation tubes and separated from the cellular components by centrifugation at 3,000 x g for 10 min, followed by separate centrifugation at 12,000 x g for 4 °C for 10 min to remove residual particulate matter. The resulting supernatant was harvested and stored at -80°C until use. Total serum RNA was extracted using 750 µL of TRizol® LS Reagent per 250 µL of serum (Thermo Fisher, Carlsbad, CA, USA) according to the reagent instructions.
PIRNA Sequence
The PIRNA sequencing library was constructed from serum RNAs from five patients (3 in TNM I-IIA and 2 in TNM III-IV) and five healthy controls followed by Illumina sequencing. Quality control was performed using FastQC, followed by the adapter trimming via a cutter pad. Sequence reads were allowed up to two discrepancies to match Pirbase using NovoAlign (v2.07.11). Differentially expressed pirna was identified as indicating | log2(Folding change) | > 2 p<0.05 based on TPM-Normalized count.
Reverse transcription and quantification by real-time PCR
Reverse transcription was performed using miRNA 1st strand cDNA synthesis kit (Accorate Biotechnology (Hunan) Co., Ltd., Changsha, China). Quantitative PCR (QPCR) analysis was performed on a LightCycler 480 system (Roche, Basel, Switzerland) using the Sybr Green Premix Pro Taq HS QPCR kit (precise biotechnology). A small U6 nuclear RNA was used as an endogenous reference gene. Relative pirna expression levels were calculated by the comparison cycle threshold (CT) method: (ΔCT=CTPirna-CTU6), as mentioned above. All QPCR primer sequences are listed in Tables 1 and S1.
ML model construction strategy
A computational framework that integrates PIRNA expression and CA153 was developed to evaluate a 10 mL algorithm for early breast cancer detection. The algorithms include K-Nearest Neighter (KNN), Logistic Regression (LR), Decision Tree (DT), Artificial Neural Network (ANN), Support Vector Machine (SVM), Gradient Boost Decision Tree (GBDT), Light Gradation Boost Machine (LGBM), Adaptive Boost (Adaboost), and Exte. All analyses use standardized libraries including Scikit-Learn (v1.5.2; https://scikit-learn.org), xgboost (v1.7.4; https://xgboost.readthethedocs.io), and lightgbm (v4.0; https:eetctgsgbm.ided prepared-fided:fidedgbm.ided prepared:fidedgbm.ided prepared:fidedgbm.ided prepared:fidedgbm. It was implemented in 3.10. The dataset consisted of 102 early stage patients and 122 healthy volunteers, which were inherently balanced. Therefore, no artificial balancing techniques were employed during model construction. The effectiveness of the model was systematically assessed using the receiver operating characteristic curve (AUC), calibration curve, and other areas under the performance metrics (sensitivity, specificity, accuracy, positive predictors). [PPV]negative predicted value [NPV]and f1 score). Clinical utility was quantified via decision curve analysis (DCA) and model interpretability was assessed using SHAP (Shapley Additive Description, 0.46.0, https://shap.readthedocs.io/en/latest/). Through this rigorous evaluation, Xgboost emerged as the best classifier. In particular, the initial model screening used a 70/30 train test split to compare classifier performances with initially identifying XGBoost as the top candidate. This selection was rigorously validated using 80/20 bootstrap-based cross-validation to ensure a robust and unbiased performance estimate.
Cross-validation was performed using a bootstrap methodology. This included 1000 iterations of resampling and resampling, generating different training sets. Each resampled dataset was split into 80% training and 20% validation subset. For each iteration, ROC curve metrics were calculated in the validation set. AUC values were collected and the true positive velocity (TPR) of aggregation was calculated as the average TPR over all iterations (NP.mean (TPRS)). Model identification performance was summarized by mean AUC, and variability was quantified by its standard deviation.
Statistical analysis
Statistical analyses were performed using SPSS software (Statistical Package for Social Sciences, 26.0, https://www.ibm.com/products/spss-statistics, cary, nc, usa) and GraphPad Prism software (9.5, https://www.graphpad.com, ca, ca, ca, ca, ca, ca). The Kolmogorov-Smirnov test was used to assess the health of the data. If normal, t-tests were used that were not used for analysis of two data groups and one-way analysis of variance (ANOVA) for analysis of multiple groups. Otherwise, the Mann-Whitney U test was used for data from two groups and for Kruskal-Wallis tests from multiple analytic groups. ROC curve analysis was performed to assess diagnostic efficacy by calculating AUC, sensitivity, specificity, and 95% confidence intervals (CIs). Statistical significance was defined as two-tailed p<0.05.
