Team:NJMU-China/Parallel Screening Strategy
Can you hear the Sound of Silence ?
NJMU-China's Web page is loading, please wait...●ω●
Bioinformatics analysis was employed to identify the specific urine metabolites of autistic children. The clinical heterogeneity of ASD calls for the need for improved sensitive and specific laboratory diagnostic methods. We used a machine-learning algorithm to find a combination of metabolites for screening. As for the output metabolites list, we would further verify according to biosensors in a high-throughput way based on systematic-level Escherichia coli genome library construction and screening.
Data Collection
Unfortunately, we are not allowed to collect urine samples from children with autism during the pandemic-controlling period. So, we used the metabolomics data published online instead.
A total of 57 children with ASD and 82 typical development (TD) children were recruited for this study1. The inclusion criteria for ASD children were: boys or girls younger than 14 years, diagnosis of ASD according to the criteria in the Diagnostic and Statistical Manual of Mental Disorders V (American Psychiatric Association, 2013). The study was conducted in two phases. The discovery cohort included 28 ASD children and 41 TD children. The validation cohort consisted of 29 ASD children and 41 TD children. Differences between urinary amino acids (UAA) and their metabolites are assessed by LS-MS/MS. And a total of 63 UAA indicators were identified including serotonin.
Statistical Analysis
Quality control includes outlier evaluation and variable transformation. Outliers were evaluated by Tukey’s test and removed if necessary in order to achieve better statistical inference. As for the skewed distribution, log-transform was used to normalized the data.
Among the 63 urine metabolites, group differences between ASD and TD were evaluated by Student’s t-test (age) and Fisher’s exact test (sex) with age and sex stratification. CART is a Gini-index-based decision tree for classification and built by the CART package. The ability of UAA indicators to distinguish between ASD and TD children was evaluated using the Wilcoxon rank-sum test.
Statistical analyses were performed by R (version 3.6.3) and RStudio (version 1.2.5033). All statistical tests were two-sided. The significance level was set at α = 0.05 and applied to all the analyses.
Results
 |
 |
A decision tree is a graphical depiction of a decision and every potential outcome or result of making that decision. Decision Tree algorithms can be applied and used in various different fields. It can be used as a replacement for statistical procedures to find data, to extract text, to find missing data in a class, to improve search engines and it also finds various applications in medical fields2. Using decision tree modeling and receiver operating characteristic curve analysis, we identified a panel of 4 UAA indicators that discriminated between the samples from ASD and TD children (5-Hydroxy-Tryptamine (serotonin), 2-aminoisobutyric acid, ethanolamine, proline). Serotonin is located at the root node, indicating that it can decrease the Gini-index the most, having the best classification efficiency. The results are consistent with previous conclusions published. The significantly altered of the four indicators could therefore be potential diagnostic biomarkers for ASD.
Reference
1. Liu, A. et al. Altered Urinary Amino Acids in Children With Autism Spectrum Disorders. 13, 1–9 (2019).
2. A. Navada, A. N. Ansari, S. Patil and B. A. Sonkamble, "Overview of use of decision tree algorithms in machine learning," 2011 IEEE Control and System Graduate Research Colloquium, Shah Alam, 2011, pp. 37-42, doi: 10.1109/ICSGRC.2011.5991826.