Tree-based ensemble machine learning models in the prediction of acute respiratory distress syndrome following cardiac surgery: a multicenter cohort study.

BACKGROUND: Acute respiratory distress syndrome (ARDS) after cardiac surgery is a severe respiratory complication with high mortality and morbidity. Traditional clinical approaches may lead to under recognition of this heterogeneous syndrome, potentially resulting in diagnosis delay. This study aims to develop and external validate seven machine learning (ML) models, trained on electronic health records data, for predicting ARDS after cardiac surgery. METHODS: This multicenter, observational cohort study included patients who underwent cardiac surgery in the training and testing cohorts (data from Nanjing First Hospital), as well as those patients who had cardiac surgery in a validation cohort (data from Shanghai General Hospital). The number of important features was determined using the sliding windows sequential forward feature selection method (SWSFS). We developed a set of tree-based ML models, including Decision Tree, GBDT, AdaBoost, XGBoost, LightGBM, Random Forest, and Deep Forest. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and Brier score. The SHapley Additive exPlanation (SHAP) techinque was employed to interpret the ML model. Furthermore, a comparison was made between the ML models and traditional scoring systems. ARDS is defined according to the Berlin definition. RESULTS: A total of 1996 patients who had cardiac surgery were included in the study. The top five important features identified by the SWSFS were chronic obstructive pulmonary disease, preoperative albumin, central venous pressure_T4, cardiopulmonary bypass time, and left ventricular ejection fraction. Among the seven ML models, Deep Forest demonstrated the best performance, with an AUC of 0.882 and a Brier score of 0.809 in the validation cohort. Notably, the SHAP values effectively illustrated the contribution of the 13 features attributed to the model output and the individual feature's effect on model prediction. In addition, the ensemble ML models demonstrated better performance than the other six traditional scoring systems. CONCLUSIONS: Our study identified 13 important features and provided multiple ML models to enhance the risk stratification for ARDS after cardiac surgery. Using these predictors and ML models might provide a basis for early diagnostic and preventive strategies in the perioperative management of ARDS patients.

[Retracted] LOXL1­AS1 promotes thymoma and thymic carcinoma progression by regulating miR­525­5p­HSPA9.

Following the publication of the above article, a concerned reader drew to the Editor's attention that certain of the cell invasion assay data featured in Figs. 2G and H, 5M and N, and 9K and L, and the tumor images shown in Fig. 6B, were strikingly similar to data appearing in different form in other articles written by different authors at different research institutes that had either already been published elsewhere prior to the submission of this paper to Oncology Reports, or were under consideration for publication at around the same time (some of which have been retracted). In view of the fact that certain of these data had already apparently been published prior to the submission of this article for publication, the Editor of Oncology Reports has decided that this paper should be retracted from the Journal. The authors were asked for an explanation to account for these concerns, but the Editorial Office did not receive a satisfactory reply. The Editor apologizes to the readership for any inconvenience caused. [Oncology Reports 45: 117, 2021; DOI: 10.3892/or.2021.8068].

Neoadjuvant camrelizumab (an anti-PD-1 antibody) plus chemotherapy or apatinib (a VEGFR-2 inhibitor) for initially unresectable stage II-III non-small-cell lung cancer: a multicentre, two-arm, phase 2 exploratory study.

This multicentre, two-arm, phase 2 study aimed to explore the efficacy and safety of neoadjuvant camrelizumab plus chemotherapy or apatinib in patients with initially unresectable stage II-III non-small-cell lung cancer (NSCLC). Eligible patients regardless of PD-L1 expression received neoadjuvant camrelizumab 200 mg and platinum-doublet chemotherapy every 3 weeks (arm A) or those with PD-L1-positive tumors received neoadjuvant camrelizumab and apatinib 250 mg once daily (arm B), for 2-4 cycles, followed by surgery. The primary endpoint was major pathological response (MPR) rate. Thirty patients in arm A and 21 in arm B were enrolled. Surgery rates were 50.0% (15/30) in arm A and 42.9% (9/21) in arm B, with all patients achieving R0 resections. Of these patients, the MPR and pathological complete response rates were both 20.0% (95% CI 4.3-48.1) in arm A and were 55.6% (95% CI 21.2-86.3) and 11.1% (95% CI 0.3-48.2) in arm B, respectively. The corresponding objective response rates were 33.3% (95% CI 11.8-61.6) and 55.6% (95% CI 21.2-86.3). With a median follow-up of 22.4 months (95% CI 19.0-26.0), the median event-free survival was not reached (NR; 95% CI 13.6-NR) in arm A and 16.8 months (95% CI 8.6-NR) in arm B. Grade 3 or above treatment-related adverse events occurred in eight (26.7%) patients in arm A and three (14.3%) in arm B. Biomarker analysis showed baseline TYROBP expression was predictive of treatment response in arm B. Neoadjuvant camrelizumab plus chemotherapy or apatinib exhibits preliminary efficacy and manageable toxicity in patients with initially unresectable stage II-III NSCLC.

A metabolomics study on carcinogenesis of ground-glass nodules.

Objective: This study aimed to identify differential metabolites and key metabolic pathways between lung adenocarcinoma (LUAD) tissues and normal lung (NL) tissues using metabolomics techniques, to discover potential biomarkers for the early diagnosis of lung cancer. Material and Methods: Forty-five patients with primary ground-glass nodules (GGN) identified on computed tomography imaging and who were willing to undergo surgery at Shanghai General Hospital from December 2021 to December 2022 were recruited to the study. All participants underwent video thoracoscopy surgery with segmental or wedge resection of the lung. Tissue samples for pathological examination were collected from the site of ground-glass nodules (GGN) lesion and 3 cm away from the lesion (NL). The pathology results were 35 lung adenocarcinoma (LUAD) cases (13 invasive adenocarcinoma, 14 minimally invasive adenocarcinoma, and eight adenocarcinoma in situ), 10 benign samples, and 45 NL tissues. For the untargeted metabolomics technique, 25 LUAD samples were assigned as the case group and 30 NL tissues as the control group. For the targeted metabolomics technique, ten LUAD samples were assigned as the case group and 15 NL tissues as the control group. Samples were analyzed by untargeted and targeted metabolomics, with liquid chromatography-tandem mass spectrometry detection used as part of the experimental procedure. Results: Untargeted metabolomics revealed 164 differential metabolites between the case and control groups, comprising 110 up regulations and 54 down regulations. The main metabolic differences found by the untargeted method were organic acids and their derivatives. Targeted metabolomics revealed 77 differential metabolites between the case and control groups, comprising 69 up regulations and eight down regulations. The main metabolic changes found by the targeted method were fatty acids, amino acids, and organic acids. The levels of organic acids such as lactic acid, fumaric acid, and malic acid were significantly increased in LUAD tissue compared to NL. Specifically, an increased level of L-lactic acid was found by both untargeted (variable importance in projection [VIP] = 1.332, fold-change [FC] = 1.678, q = 0.000) and targeted metabolomics (VIP = 1.240, FC = 1.451, q = 0.043). Targeted metabolomics also revealed increased levels of fumaric acid (VIP = 1.481, FC = 1.764, q = 0.106) and L-malic acid (VIP = 1.376, FC = 1.562, q = 0.012). Most of the 20 differential fatty acids identified were downregulated, including dodecanoic acid (VIP = 1.416, FC = 0.378, q = 0.043) and tridecane acid (VIP = 0.880, FC = 0.780, q = 0.106). Furthermore, increased levels of differential amino acids were found in LUAD samples. Conclusion: Lung cancer is a complex and heterogeneous disease with diverse genetic alterations. The study of metabolic profiles is a promising research field in this cancer type. Targeted and untargeted metabolomics revealed significant differences in metabolites between LUAD and NL tissues, including elevated levels of organic acids, decreased levels of fatty acids, and increased levels of amino acids. These metabolic features provide valuable insights into LUAD pathogenesis and can potentially serve as biomarkers for prognosis and therapy response.

Evaluating personalized circulating tumor DNA detection for early-stage lung cancer.

Circulating tumor DNA (ctDNA) has been widely used as a minimally invasive biomarker in clinical routine. However, a number of factors such as panel design, sample quality, patients' disease stages are known to influence ctDNA detection sensitivity. In this study, we systematically evaluated common factors associated with the variability of ctDNA detection in plasma and investigated ctDNA abundance in bronchoalveolar lavage (BAL). Whole exome profiling was conducted on 61 tumor tissue samples to identify tumor-specific variants, which were then used to design personalized assay MarRyDa® for ctDNA detection. DNA extracted from BAL fluid and plasma were genotyped using MarRyDa® platform. Our analysis showed that histological subtypes and disease stages had significant differences in ctDNA detection rate. Furthermore, we found that DNA purified from BAL supernatants contains the highest levels of ctDNA compared with BAL precipitates and plasma; therefore, utilizing BAL supernatants for tumor detection might provide additional benefits. Finally, we demonstrated that tumor cellularity played significant roles in the design of personalized ctDNA panel which eventually impacts ctDNA detection sensitivity. We suggest setting a flexible criteria for sample quality control and utilization of BAL might benefit more patients in clinics.