Machine Learning Models for Genetic Risk Assessment of Infants with Non-Syndromic Orofacial Cleft

Abstract

The isolated type of orofacial cleft, termed non-syndromic cleft lip with or without cleft palate (NSCL/P), is the second most common birth defect in China, with Asians having the highest incidence in the world. NSCL/P involves multiple genes and complex interactions between genetic and environmental factors, imposing difficulty for the genetic assessment of the unborn fetus carrying multiple NSCL/P-susceptible variants. Although genome-wide association studies (GWAS) have uncovered dozens of single nucleotide polymorphism (SNP) loci in different ethnic populations, the genetic diagnostic effectiveness of these SNPs requires further experimental validation in Chinese populations before a diagnostic panel or a predictive model covering multiple SNPs can be built. In this study, we collected blood samples from control and NSCL/P infants in Han and Uyghur Chinese populations to validate the diagnostic effectiveness of 43 candidate SNPs previously detected using GWAS. We then built predictive models with the validated SNPs using different machine learning algorithms and evaluated their prediction performance. Our results showed that logistic regression had the best performance for risk assessment according to the area under curve. Notably, defective variants in MTHFR and RBP4, two genes involved in folic acid and vitamin A biosynthesis, were found to have high contributions to NSCL/P incidence based on feature importance evaluation with logistic regression. This is consistent with the notion that folic acid and vitamin A are both essential nutritional supplements for pregnant women to reduce the risk of conceiving an NSCL/P baby. Moreover, we observed a lower predictive power in Uyghur than in Han cases, likely due to differences in genetic background between these two ethnic populations. Thus, our study highlights the urgency to generate the HapMap for Uyghur population and perform resequencing-based screening of Uyghur-specific NSCL/P markers.

Machine Learning Algorithms for Predicting the Recurrence of Stage IV Colorectal Cancer After Tumor Resection

“…In the present study, four basic ML algorithms-logical regression 9 , decision tree 10 , GradientBoosting 11 and lightGBM 4,12 -are implemented 13,14 . Logistic regression is a classical classification method in statistical learning.…”

Radiomics and Machine Learning in Oral Healthcare

“…Consequently, automatic methods based on deep neural networks have been tested for several purposes, which are as follows: classification, image registration, segmentation, lesion detection, image retrieval, image guided therapy, image generation, and enhancement . Most recently, radiomics and AI research have been advancing in the dental field, revealing the potential of these technologies to substantially improve clinical care …”

Environmental mechanisms of orofacial clefts

“…In human populations, there is some evidence for polymorphisms of genes associated with RA metabolism and signaling that could increase the risk for OFCs. A study on Han and Uyghur populations in China found an association between offspring RBP4 polymorphism and increased risk for CL/P using a machine learning model (S. J. Zhang et al, 2018). Two other studies in Chinese populations and a study in an Indian population found correlation between RARA polymorphism and increased risk for CL/P (Fan, Li, & Wu, 2007; Peanchitlertkajorn, Cooper, Liu, Field, & Marazita, 2003; Xavier et al, 2013).…”