BACKGROUND: Neonatal acne occurs in the first few weeks after birth. Some lesions are more serious and leave scars. Maternal surface skin lipids (SSL) have a strong correlation with SSL of infants. The establishment of prediction rank model based on maternal SSL is essential to the prevention and treatment of neonatal acne. METHOD: Surface skin lipids samples were collected from the mothers (M) of 56 neonatal acne patients and the mothers (HM) of 19 healthy infants. Surface skin lipids from the right forehead were collected using a noninvasive method. UPLC-QTOF-MS was applied to detect SSL. Partial least squares discriminant analysis and receiver operating characteristic (ROC) analysis were performed to screen and validate potential lipids. Random forest (RF) and ROC analysis were used to establish a prediction model and evaluate its accuracy. RESULTS: Sixteen altered potential lipids belonging to fatty acids, sphingomyelins, and glycerides were associated with M. M had less lipids than HM. Spearman's correlation of 16 lipids revealed 9 with high correlation. They were chosen as characteristic values of the RF prediction model. And the model showed an average accuracy of 98% in the validation set. CONCLUSION: We have established an RF model for predicting neonatal acne and have shown that high skin barrier-related lipids were markers for predicting neonatal acne.
BACKGROUND:Neonatal acne occurs in the first few weeks after birth. Some lesions are more serious and leave scars. Maternal surface skin lipids (SSL) have a strong correlation with SSL of infants. The establishment of prediction rank model based on maternal SSL is essential to the prevention and treatment of neonatal acne. METHOD: Surface skin lipids samples were collected from the mothers (M) of 56 neonatal acnepatients and the mothers (HM) of 19 healthy infants. Surface skin lipids from the right forehead were collected using a noninvasive method. UPLC-QTOF-MS was applied to detect SSL. Partial least squares discriminant analysis and receiver operating characteristic (ROC) analysis were performed to screen and validate potential lipids. Random forest (RF) and ROC analysis were used to establish a prediction model and evaluate its accuracy. RESULTS: Sixteen altered potential lipids belonging to fatty acids, sphingomyelins, and glycerides were associated with M. M had less lipids than HM. Spearman's correlation of 16 lipids revealed 9 with high correlation. They were chosen as characteristic values of the RF prediction model. And the model showed an average accuracy of 98% in the validation set. CONCLUSION: We have established an RF model for predicting neonatal acne and have shown that high skin barrier-related lipids were markers for predicting neonatal acne.