RESUMO
BACKGROUND & AIMS: We previously identified subsets of patients with metabolic (dysfunction)-associated steatotic liver disease (MASLD) with different metabolic phenotypes. Here, we aimed to refine this classification based on genetic algorithms implemented in a Python package. The use of these genetic algorithms can help scientists to solve problems which cannot be solved with other methods. We present this package and its capabilities with specific problems. The name, PyGenMet, comes from its main goal, solving problems in Python with Genetic Algorithms and Metabolomics data. METHODS: We collected serum from methionine adenosyltransferase 1a knockout (Mat1a-KO) mice, which have chronically low level of hepatic S-adenosylmethionine (SAMe) and the metabolomes of all samples were determined. We also analyzed serum metabolomes of 541 patients with biopsy proven MASLD (182 with simple steatosis and 359 with metabolic (dysfunction)-associated steatohepatitis or MASH) and compared them with the serum metabolomes of this specific MASLD mouse model using Genetic Algorithms in order to select patients with a specific phenotype. RESULTS: By applying genetic algorithms, we have found a subgroup of patients with a lipid profile similar to that observed in the mouse model. When analyzing the two groups of patients, we have seen that patients with a lipid profile reflecting the mouse model characteristics show significant differences in lipoproteins, especially in LDL-4, LDL-5, and LDL-6 associated with atherogenic risk. CONCLUSION: The results show that the application of genetic algorithms to subclassify patients with MASLD (or other metabolic disease) give consistent results and are a good approximation for the treatment of large volumes of data such as those from omics sciences and patient classification.
RESUMO
BACKGROUND AND AIM: Type 2 Diabetes mellitus (T2DM), age, and obesity are risk factors for metabolic dysfunction-associated steatotic liver disease (MASLD). We aimed to assess the performance of non-invasive tests (NITs) for the diagnosis of metabolic dysfunction-associated steatohepatitis (MASH) and fibrosis in high-risk subjects. METHODS: Multicentre cross-sectional study that included 124 biopsy-proven MASLD in more than 50 years-old patients with overweight/obesity and T2DM. Vibration-controlled transient elastography, Fibrosis-4 index (FIB-4), Non-alcoholic fatty liver disease fibrosis score (NFS), OWLiver Panel (OWLiver DM2 + Metabolomics-Advanced Steatohepatitis Fibrosis Score -MASEF) and FibroScan-AST were performed. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the receiver operating characteristic curve (AUC) were calculated. NITs were assessed individually and in sequential/parallel combinations. RESULTS: 35 (28.2%) patients had early MASH and 66 (53.2%) had MASH with significant fibrosis (at-risk MASH). The OWLiver Panel correctly classified 86.1% as MASH, showing an accuracy, sensitivity, specificity, PPV, and NPV of 0.77, 0.86, 0.35, 0.85, and 0.36, respectively. Class III obesity, diabetes control, or gender did not impact on the performance of the OWLiver Panel (p > 0.1). NITs for at-risk MASH showed an AUC > 0.70 except for NFS. MASEF showed the highest accuracy and NPV for at-risk MASH (AUC 0.77 [0.68-0.85], NPV 72%) and advanced fibrosis (AUC 0.80 [0.71-0.88], NPV 92%). Combinations of NITs for the identification of at-risk MASH did not provide any additional benefit over using MASEF alone. CONCLUSION: One-step screening strategy with the OWLiver Panel has high accuracy to detect MASH and at-risk MASH in high-risk subjects for MASLD.