Predicting the diagnostic efficacy of trio-based whole exome sequencing in children with low-function autism spectrum disorders: a multicenter study.

Background: Although significant progress has been made in trio-based whole-exome sequencing (trio-WES) that enables the detection of exon-level variants, the diagnostic effectiveness of empirical and unselected use of trio-WES in children with low-function autism spectrum disorders (LF-ASDs) remains unsatisfactory. Thus, the identification of an appropriate approach for predicting the diagnostic efficacy of trio-WES at the pre-diagnosis stage is essential for implementing individualized diagnosis for children with LF-ASDs. Methods: A total of 168 LF-ASDs patients who underwent trio-WES at Sun Yat-sen Memorial Hospital from September 2016 to December 2022 were enrolled as the training set. Additionally, 58 LF-ASDs patients who received trio-WES at Weierkang Children's Rehabilitation Center between January 2023 and December 2023 were recruited as an independent external validation set. Univariate and multivariate binary logistic analyses were performed on the training set to select phenotypic variables to establish a nomogram. The discriminative performance of the model was evaluated using receiver operating characteristic (ROC) curves and calibration curves. Furthermore, the nomogram was validated in external validation sets. Results: Univariate and multivariate analyses identified independent trio-WES diagnosis-related predictive indicators, including severity of global developmental delay/intellectual disability, complexity of neurodevelopmental/neurological comorbid conditions, head circumference abnormalities, and brain malformations, in the training cohort and used to develop a nomogram. The nomogram showed excellent discrimination performance, with an area under curve (AUC) of the ROC in the training cohort of 0.868 (95% CI: 0.811-0.925), resulting in sensitivity, specificity, accuracy, precision, and F1 score values of 85.56, 82.05, 83.93, 84.62%, and 0.85, respectively. The model also exhibited strong prediction ability in the external validation set (AUC: 0.941, 95% CI: 0.880-0.998; sensitivity: 85.29%; specificity: 91.67%; accuracy: 87.93%; precision: 93.55%; and F1 score: 0.89). Moreover, the calibration curves demonstrated good agreement between the nomogram predictions and actual observations in both training and validation sets. Conclusion: We developed an user-friendly and highly accurate model for predicting the diagnostic probability of trio-WES in LF-ASDs children, which could help implement an individualized diagnostic strategy for affected children and their families at the pre-diagnosis stage.