Modelling of SNP markers' features related to nose morphology

In forensic science, single nucleotide polymorphism (SNP) markers can be used to identify, phenotype, determine lineage, and determine genetic lineage. Forensic phenotyping has become an important tool thanks to advances in molecular biology and genetics. This method provides valuable information in crime solving, identification, and prehistoric human research. In recent years, studies on various physical characteristics such as eye, hair, and skin color, male pattern baldness, facial structure, ear shape, and height have increased significantly. Genetic factors that determine facial structure are determined by the interaction of many genes. These genes affect features such as face width, length, nose shape, and prominence of cheekbones. Large-scale genome studies conducted in recent years have enabled the identification of important genetic variations affecting facial structure. In this study, two multiplex panels consisting of 24-SNP markers that were determined to be highly significant in predicting nasal morphology were developed and optimized. 2800M control DNA was used in the optimization process. Optimization of the panels was completed and analysis threshold, sensitivity, dynamic range and repeatability validation parameters were applied. Oral swab or blood samples were taken from 103 volunteers aged 18-45 years who had no chronofacial anomalies in the facial region and had not undergone facial surgery. SNP markers associated with the nose were analyzed with the SNaPshot method and a prediction model was performed with Multiple Nominal Logistic Regression (MLR) analysis using the Python programming language (ML, Machine Learning) for the prediction of nasal features. The model was created from 24 SNP locus genotype data and 5 target variables from nasal features (long nose, wide nostril, long nostril, arched nose and wide nose). Data for each target variable were divided into training and test sets (70% training, 30% test). After training the MLR model using the training set, nasal morphology was predicted on the test set. Finally, accuracy, precision, sensitivity, F1 score, ROC curve and AUC values were obtained for the model. According to the prediction evaluation made based on AUC values showing the performance of the models created for 5 nasal features; The highest prediction for long nose (AUC = 0.58) was rs11665450 (TT) genotype. The highest prediction for wide nose (AUC = 0.42) was rs2045323 (GA) SNP genotype. The highest prediction for nasal bridge angle (arched, non-arched) was shown by the SNP locus with the rs921120 (GA) genotype (AUC = 0.69). The highest prediction for wide nostril (AUC = 0.73) was rs2206437 (TT) genotype. The SNP locus with the genotype rs11653132 (AC) showed the highest prediction (AUC = 0.70) for long nostril. As a result of our study, it was seen that wide nostril (rs921120), nasal bridge (rs11653132), long nostril (rs2206437), long nose (rs11665450), wide nose (rs2045323) models performed better with balanced true positive and true negative rates and higher AUC values. However, in addition to these SNP markers, more SNP loci related to nasal morphology can be investigated and modelled with higher accuracy.