Necrotizing enterocolitis (NEC) is a life-threatening condition for premature infants in neonatal intensive care units. Finding indicators that can predict NEC development before symptoms appear would provide more time to apply targeted interventions. In this study, stools from 132 very-low-birth-weight (VLBW) infants were collected daily in the context of a multi-center prospective study aimed at investigating the potential of fecal biomarkers for NEC prediction using proteomics technology. Eight of the VLBW infants received a stage-3 NEC diagnosis. Stools collected from the NEC infants up to 10 days before their diagnosis were available for seven of them. Their samples were matched with those from seven pairs of non-NEC controls. The samples were processed for liquid chromatography-tandem mass spectrometry analysis using SWATH/DIA acquisition and cross-compatible proteomic software to perform label-free quantification. ROC curve and principal component analyses were used to explore discriminating information and to evaluate candidate protein markers. A series of 36 proteins showed the most efficient capacity with a signature that predicted all seven NEC infants at least a week in advance. Overall, our study demonstrates that multiplexed proteomic signature detection constitutes a promising approach for the early detection of NEC development in premature infants.
Inflammatory bowel disease (IBD) flare-ups exhibit symptoms that are similar to other diseases and conditions, making diagnosis and treatment complicated. Currently, the gold standard for diagnosing and monitoring IBD is colonoscopy and biopsy, which are invasive and uncomfortable procedures, and the fecal calprotectin test, which is not sufficiently accurate. Therefore, it is necessary to develop an alternative method. In this study, our aim was to provide proof of concept for the application of Sequential Window Acquisition of All Theoretical Mass Spectra-Mass spectrometry (SWATH-MS) and machine learning to develop a non-invasive and accurate predictive model using the stool proteome to distinguish between active IBD patients and symptomatic non-IBD patients. Proteome profiles of 123 samples were obtained and data processing procedures were optimized to select an appropriate pipeline. The differentially abundant analysis identified 48 proteins. Utilizing correlation-based feature selection (Cfs), 7 proteins were selected for proceeding steps. To identify the most appropriate predictive machine learning model, five of the most popular methods, including support vector machines (SVMs), random forests, logistic regression, naive Bayes, and k-nearest neighbors (KNN), were assessed. The generated model was validated by implementing the algorithm on 45 prospective unseen datasets; the results showed a sensitivity of 96% and a specificity of 76%, indicating its performance. In conclusion, this study illustrates the effectiveness of utilizing the stool proteome obtained through SWATH-MS in accurately diagnosing active IBD via a machine learning model.
2 October 2025 in Clinical and Translational Gastroenterology
INTRODUCTION: Crohn's disease (CD) and ulcerative colitis (UC) have overlapping symptoms, but they differ in pathology and treatment. Currently, distinguishing between these diseases involves invasive procedures such as colonoscopy and histopathology. Fecal proteins, stable and in direct contact with inflammation, offer a noninvasive alternative. This study focuses on using high-throughput data-independent acquisition mass spectrometry and machine learning to develop an accurate biomarker signature from complex stool samples. METHODS: Stool samples obtained from 69 active patients were analyzed. Analysis of the stool proteome led to the identification and quantification of approximately 1,250 proteins. The samples were divided into training and testing groups. After data processing, various feature selection algorithms were applied on the training group to determine proteins that were significantly different between the CD and UC groups. In addition, 6 machine learning algorithms were evaluated to identify the best-performing classifiers. RESULTS: Sixteen proteins were selected based on several feature selection algorithms, and 6 models were trained based on them. According to the performance metrics of each algorithm on the training data set, the Naive Bayes model was selected. For performance validation, the final predictive model was applied to 16 blind prospective samples as the test data set. Notably, the model achieved an area under the curve of 0.96 on both the training and test data sets, highlighting its robustness and stability. DISCUSSION: This study demonstrates the potential of combining multiple stool protein biomarkers through high-throughput data-independent acquisition mass spectrometry and machine learning tools to develop a predictive model for efficiently distinguishing CD from UC.