Developing a peripheral blood RNA-seq based NETseq ensemble classifier: A potential novel tool for non-invasive detection and treatment response assessment in neuroendocrine tumor patients receiving 177 Lu-DOTATATE PRRT.

Autor: Padwal MK; Molecular Biology Division, Bhabha Atomic Research Centre, Mumbai, India.; Homi Bhabha National Institute, Mumbai, India., Parghane RV; Homi Bhabha National Institute, Mumbai, India.; Radiation Medicine Centre, Bhabha Atomic Research Centre, Tata Memorial Centre Annexe, Mumbai, India., Chakraborty A; Homi Bhabha National Institute, Mumbai, India.; Radiation Medicine Centre, Bhabha Atomic Research Centre, Tata Memorial Centre Annexe, Mumbai, India., Ujaoney AK; Molecular Biology Division, Bhabha Atomic Research Centre, Mumbai, India., Anaganti N; Molecular Biology Division, Bhabha Atomic Research Centre, Mumbai, India.; Homi Bhabha National Institute, Mumbai, India., Basu S; Homi Bhabha National Institute, Mumbai, India.; Radiation Medicine Centre, Bhabha Atomic Research Centre, Tata Memorial Centre Annexe, Mumbai, India., Basu B; Molecular Biology Division, Bhabha Atomic Research Centre, Mumbai, India.; Homi Bhabha National Institute, Mumbai, India.
Jazyk: angličtina
Zdroj: Journal of neuroendocrinology [J Neuroendocrinol] 2024 Nov 13, pp. e13462. Date of Electronic Publication: 2024 Nov 13.
DOI: 10.1111/jne.13462
Abstrakt: Neuroendocrine tumors (NETs) are presented with metastases due to delayed diagnosis. We aimed to identify NET-related biomarkers from peripheral blood. The development and validation of a multi-gene NETseq ensemble classifier using peripheral blood RNA-Seq is reported. RNA-Seq was performed on peripheral blood samples from 178 NET patients and 73 healthy donors. Distinguishing gene features were identified from a learning cohort (59 PRRT-naïve GEP-NET patients and 38 healthy donors). Ensemble classifier combining the output of five machine learning algorithms viz. Random Forest (RF), Extreme Gradient Boosting (XGBOOST), Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and Logistic Regression (LR) were trained and independently validated in the evaluation cohort (n = 106). The response to PRRT was evaluated in the PRRT cohort (n = 46) and the PRRT response monitoring cohort (n = 16). The response to 177 Lu-DOTATATE PRRT was assessed using RECIST 1.1 criteria. The Ensemble classifier trained on 61 gene features, distinguished NET from healthy samples with 100% accuracy in the learning cohort. In an evaluation cohort, the classifier achieved 93% sensitivity (95% CI: 87.8%-98.03%) and 91.4% specificity (95% CI: 82.1%-100%) for PRRT-naïve GEP-NETs (AUROC = 95.4%). The classifier returned >87.5% sensitivity across different tumor characteristics and outperformed serum Chromogranin A sensitivity (χ 2  = 21.89, p = 4.161e-6). In the PRRT cohort, RECIST 1.1 responders showed significantly lower NETseq prediction scores after 177 Lu-DOTATATE PRRT, in comparison to the non-responders. In an independent response monitoring cohort, paired samples (before PRRT and after 2nd or 3rd cycle of PRRT) were analyzed. The NETseq prediction score significantly decreased in partial responders (p = .002) and marginally reduced in stable disease (p = .068). The NETseq ensemble classifier identified PRRT-naïve GEP-NETs with high accuracy (≥92%) and demonstrated a potential role in early treatment response monitoring in the PRRT setting. This blood-based, non-invasive, multi-analyte molecular method could be developed as a valuable adjunct to conventional methods in the detection and treatment response assessment in NET patients.
(© 2024 The Author(s). Journal of Neuroendocrinology published by John Wiley & Sons Ltd on behalf of British Society for Neuroendocrinology.)
Databáze: MEDLINE
Externí odkaz: