1290-P: Gut Microbiota in New-Onset Pediatric Patients with Type 1 Diabetes: Machine Learning Algorithms to Classify Microorganisms Disease-Linked

Abstract Aims The purpose of this work is to find the gut microbial fingerprinting of pediatric patients with type 1 diabetes. Methods The microbiome of 31 children with type 1 diabetes at onset and of 25 healthy children was determined using multiple polymorphic regions of the 16S ribosomal RNA. We performed machine-learning analyses and metagenome functional analysis to identify significant taxa and their metabolic pathways content. Results Compared with healthy controls, patients showed a significantly higher relative abundance of the following most important taxa: Bacteroides stercoris, Bacteroides fragilis, Bacteroides intestinalis, Bifidobacterium bifidum, Gammaproteobacteria and its descendants, Holdemania, and Synergistetes and its descendants. On the contrary, the relative abundance of Bacteroides vulgatus, Deltaproteobacteria and its descendants, Parasutterella and the Lactobacillus, Turicibacter genera were significantly lower in patients with respect to healthy controls. The predicted metabolic pathway more associated with type 1 diabetes patients concerns “carbon metabolism,” sugar and iron metabolisms in particular. Among the clinical variables considered, standardized body mass index, anti-insulin autoantibodies, glycemia, hemoglobin A1c, Tanner stage, and age at onset emerged as most significant positively or negatively correlated with specific clusters of taxa. Conclusions The relative abundance and supervised analyses confirmed the importance of B stercoris in type 1 diabetes patients at onset and showed a relevant role of Synergistetes and its descendants in patients with respect to healthy controls. In general the robustness and coherence of the showed results underline the relevance of studying the microbioma using multiple polymorphic regions, different types of analysis, and different approaches within each analysis.

Download Full-text

Arginine metabolism and gut microbiota changes in pediatric patients with type 1 diabetes

Endocrine Abstracts ◽

10.1530/endoabs.70.aep391 ◽

2020 ◽

Author(s):

Elena Aghajanova ◽

Arthur Melkonyan ◽

Nina Alchujyan ◽

Bayburdyan Gayane ◽

Margarita Hovhannisyan ◽

...

Keyword(s):

Type 1 Diabetes ◽

Gut Microbiota ◽

Pediatric Patients ◽

Arginine Metabolism

Download Full-text

Gut Microbiota Assessment in Obese Children and Adolescents by Machine Learning Algorithms

Journal of the Endocrine Society ◽

10.1210/jendso/bvab048.1458 ◽

2021 ◽

Vol 5 (Supplement_1) ◽

pp. A716-A716

Author(s):

Giuseppe d’Annunzio ◽

Roberto Biassoni ◽

Eddi Di Marco ◽

Alberto La Valle ◽

Gianluca Piccolo ◽

...

Keyword(s):

Machine Learning ◽

Metabolic Syndrome ◽

Gut Microbiota ◽

Children And Adolescents ◽

Learning Algorithms ◽

Machine Learning Algorithms ◽

Microbiota Composition ◽

Obese Children ◽

Ruminococcus Flavefaciens ◽

Gut Microbiota Composition

Abstract Gut microbiota has been recently established to play a contributory role in the development and progression of obesity, a multifactorial disease predisposing to several comorbidities. Our aim was to evaluate the gut microbiota composition by machine learning algorithms in 33 Italian obese children and adolescents. Patients were divided in 3 groups: simple obesity (n=10, mean age 11.6 +3.0, median 10.8), metabolic syndrome (n=16, mean age 13.3+3.0, median 13.5) or Prader Willi syndrome (n=7, mean age 8.3+5.3, median 8.7). Inclusion criteria were living in Northern Italy, born singleton birth, personal history negative for acute or chronic gastrointestinal diseases and/or antibiotic or probiotics administration in the previous month. As controls 17 healthy control (mean age 12.0+2.4 median 10.6) were analyzed using the same approach. Statistical analysis for sparse high-throughput sequencing data algorithm (metagenomeSeq) using cumulative sum scaling for data normalization was performed. False discovery rate adjusted p-value using zero-inflated Gaussian fit statistical model (indicated with q). Over all analyses Parasutterella resulted with a q=0.014424, the comparison between obese patients and controls was q=0.021194. In the overall analysis Acidaminococcus intestini showed q=0.039528 while the comparison in pairs of two between metabolic syndrome and controls was q=0.03979. Using the EdgeR algorithm Clostridium bartlettii abundance between Prader Willi patients and controls resulted in q=0.02189. In overall analysis Ruminococcus flavefaciens resulted q=6.1528E-17 (using the DESeq2 algorithm); in pairs analysis Ruminococcus flavefaciens showed significant difference in Prader Willi patients as compared to obese (q=0.013755) and metabolic syndrome ones (q=0.021898). In overall analysis Veillonellaceae showed a FDR q=0.029303. while its richness resulted more than 150 times higher in metabolic syndrome patients than in controls (q=0.039793 evaluated with DESeq2 algorithm). Among Veillonellaceae descendants, the Veillonella rogosae showed, on the contrary, the lowest abundance in metabolic syndrome patients as compared to other groups. In detail, Veillonella rogosae abundances were 13 (FDR q=0.014566), around 20 times (FDR q=0.010646) and >20 (FDR q=0.0025008) less abundant in metabolic syndrome patients than obese, Prader Willi patients and controls, respectively. Significant differences in gut microbiota composition was found among patients with different degrees of obesity and controls. Further, Prader Willi patients showed a peculiar microbiota assessment.

Download Full-text

Prediction and prevention of hypoglycaemic events in type-1 diabetic patients using machine learning

Health Informatics Journal ◽

10.1177/1460458219850682 ◽

2019 ◽

Vol 26 (1) ◽

pp. 703-718 ◽

Cited By ~ 4

Author(s):

Josep Vehí ◽

Iván Contreras ◽

Silvia Oviedo ◽

Lyvia Biagi ◽

Arthur Bertachi

Keyword(s):

Machine Learning ◽

Type 1 Diabetes ◽

Blood Glucose ◽

Diabetes Management ◽

Machine Learning Algorithms ◽

Support Vector ◽

Diabetic Patients ◽

Glucose Levels ◽

Prediction And Prevention

Tight blood glucose control reduces the risk of microvascular and macrovascular complications in patients with type 1 diabetes. However, this is very difficult due to the large intra-individual variability and other factors that affect glycaemic control. The main limiting factor to achieve strict control of glucose levels in patients on intensive insulin therapy is the risk of severe hypoglycaemia. Therefore, hypoglycaemia is the main safety problem in the treatment of type 1 diabetes, negatively affecting the quality of life of patients suffering from this disease. Decision support tools based on machine learning methods have become a viable way to enhance patient safety by anticipating adverse glycaemic events. This study proposes the application of four machine learning algorithms to tackle the problem of safety in diabetes management: (1) grammatical evolution for the mid-term continuous prediction of blood glucose levels, (2) support vector machines to predict hypoglycaemic events during postprandial periods, (3) artificial neural networks to predict hypoglycaemic episodes overnight, and (4) data mining to profile diabetes management scenarios. The proposal consists of the combination of prediction and classification capabilities of the implemented approaches. The resulting system significantly reduces the number of episodes of hypoglycaemia, improving safety and providing patients with greater confidence in decision-making.

Download Full-text