Fig. 1

An overview of the ML process and potential applications to IDDs. Various types of data (e.g., clinical, behavior, neuroimaging, and multi-omics) are usually recorded in IDD cohorts. These data sets are first individually processed and cleaned to remove noise and extract relevant biological signals (feature extraction). Then, an AI/ML algorithm is trained to find rules and patterns in the integrated dataset. The choice of the algorithm usually depends on the formulation of the biological problem and other data-set specific factors (discussed in Machine learning methods and applications to IDDs section of the main text). Typically, the model can be tested objectively with independent datasets or prior knowledge. A correctly evaluated and validated model is often generalizable, and such models have a variety of clinical and laboratory applications in IDDs