Biochemistry and biophysics reports · 2026
Urinary extracellular vesicle microRNA profiles and prostate cancer biopsy outcome: Moderate discrimination and quantified optimism under leakage-controlled machine learning.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Slalmi Ayoub, Rabbah Nabila, Debbarh Ikram, Battas Ilham, Medromi Hicham, Abourriche Abdelmjid
Аннотация
BACKGROUND: Urinary extracellular-vesicle-associated microRNAs are candidates for non-invasive prostate cancer assessment. We evaluated discrimination between GSE159177 records annotated with a Gleason Score and those labelled Control; individual negative-biopsy status could not be confirmed.
RESULTS: The frozen analysis set comprised 465 urinary sample records (278 Gleason-labelled; 187 Control-labelled). Biological annotation retained 2578 mature human microRNA probes from 6658 array probes. Five-fold nested cross-validation confined feature selection, scaling, model-family selection and threshold estimation to outer-training data, yielding one held-out prediction per record. Across five repetitions, mean pooled area under the receiver-operating-characteristic curve (AUC) was 0.707 (sample standard deviation 0.017; range 0.681-0.726). The original seed-42 AUC was 0.726 (95% confidence interval 0.680-0.772), the highest repetition estimate. Its thresholded predictions gave 55.8% sensitivity and missed 123 of 278 positive-labelled records (44.2%). Global outcome-informed feature selection increased AUC by 0.041 (paired 95% confidence interval 0.003-0.080) in the matched experiment, quantifying one source of optimism. hsa-let-7b-5p was the only Mann-Whitney result significant after Benjamini-Hochberg correction and was recurrently selected, but remains exploratory.
CONCLUSIONS: This analysis combines biological annotation, leakage-controlled nested evaluation and an empirical leakage comparison to inform independent urinary replication. Discrimination was moderate and partition-dependent. Participant independence remained unresolved without a grouping key; repetitions reused the same records and did not constitute independent validation. Limited sensitivity and imperfect calibration preclude clinical use. Targeted experimental confirmation and participant-level validation are required before clinical translation. Incremental clinical value and economic benefits were not evaluated.
RESULTS: The frozen analysis set comprised 465 urinary sample records (278 Gleason-labelled; 187 Control-labelled). Biological annotation retained 2578 mature human microRNA probes from 6658 array probes. Five-fold nested cross-validation confined feature selection, scaling, model-family selection and threshold estimation to outer-training data, yielding one held-out prediction per record. Across five repetitions, mean pooled area under the receiver-operating-characteristic curve (AUC) was 0.707 (sample standard deviation 0.017; range 0.681-0.726). The original seed-42 AUC was 0.726 (95% confidence interval 0.680-0.772), the highest repetition estimate. Its thresholded predictions gave 55.8% sensitivity and missed 123 of 278 positive-labelled records (44.2%). Global outcome-informed feature selection increased AUC by 0.041 (paired 95% confidence interval 0.003-0.080) in the matched experiment, quantifying one source of optimism. hsa-let-7b-5p was the only Mann-Whitney result significant after Benjamini-Hochberg correction and was recurrently selected, but remains exploratory.
CONCLUSIONS: This analysis combines biological annotation, leakage-controlled nested evaluation and an empirical leakage comparison to inform independent urinary replication. Discrimination was moderate and partition-dependent. Participant independence remained unresolved without a grouping key; repetitions reused the same records and did not constitute independent validation. Limited sensitivity and imperfect calibration preclude clinical use. Targeted experimental confirmation and participant-level validation are required before clinical translation. Incremental clinical value and economic benefits were not evaluated.