AI-Driven Detection of Inherited Neurological Disorders Using Genomic and Multi-Omics Data

Abstract

Advances in Genomics and Multi-Omics Technology have opened new opportunities to understand the molecular basis for genetic forms of neurological disorders than ever before, the scale (size), heterogeneity (differences), and complexity of these databases create challenges for analytical approaches. To meet these challenges, Artificial Intelligence (AI) and Machine Learning (ML) have emerged as effective tools for integrating and analyzing large datasets across different omics. This systematic review summarises current evidence related to AI-based methods for detecting, classifying, and interpreting inherited forms of neurological disease using Multi-Omics data. The review demonstrates that the use of AI algorithms, especially Deep Learning and Graph-Based Frameworks, consistently outperforms traditional statistical methods in identifying disease-causing genetic variants (variant prioritization), predicting disease risk (disease prediction), and determining the relationship between genotype and phenotype (genotype-phenotype association). Integration of different omics was shown to significantly improve diagnostic accuracy and biological understanding when compared to using only one type of Omics, thereby allowing for better disease stratification (categorization) and earlier detection. Explainable AI techniques are being increasingly incorporated into models to provide transparency (i.e., explanation) for the model output and improve the ability of clinicians to interpret the output given their own context, which is a significant barrier for the translational implementation of these models. Nonetheless, there remain significant limitations, such as small sample sizes for many rare genetic disorders, the heterogeneous nature of the data (i.e., differences in the way data is collected and processed) across platforms, limited external validation of the results, and ethical issues related to data privacy and algorithmic bias. Nevertheless, critical limitations persist, including small sample sizes for rare disorders, data heterogeneity across platforms, limited external validation, and ethical concerns related to data privacy and algorithmic bias. This review highlights emerging solutions, such as federated learning, standardized data harmonization pipelines, and clinically guided model development, as pathways toward more robust and generalizable applications. Overall, AI-enabled integration of genomic and multi-omics data represents a promising paradigm for precision neurology, with the potential to shorten diagnostic timelines, improve risk prediction, and support personalised therapeutic strategies for inherited neurological disorders.