Artificial intelligence (AI) and machine learning (ML) are rapidly growing and becoming essential research tools in the world of science. Genetics is no exception, and many approaches have been developed to incorporate a variety of these techniques. For example, Support Vector Machine (SVM) was used for genome-wide automatic variant filtering in the early phase of the 1000 genomes project, or for integrating prediction scores of the variant impacts, such as SIFT or PolyPhen-2, through CADD (Combined Annotation-Dependent Depletion).

The convolutional neural network, which revolutionized image learning and initially led the third wave of the AI boom, also forms the basis of several state-of-the-art genetic methods. These include DeepVariant for variant calling, SpliceAI for predicting the variant impact on splicing patterns, and DeepSEA, ExPecto, and Basenji for predicting the effect of non-coding genetic variations on gene expression levels.

Lastly, a language model with contextual understanding, based on Transformer architecture, realized highly accurate protein structure prediction by AlphaFold, followed by a better prediction of variant impacts with fine-tuning by human and primate frequency data, as seen by AlphaMissense. The Transformer-based language model was further utilized for predicting non-coding variant functions by Enformer. In addition, generative AI based on language models (e.g., large-scale language model) is being used to support diagnosis of rare diseases.

As shown by these examples, the application of AI techniques in genetics is one of the highly productive research fields that has explored many new possibilities. It is also worth noting that AI has significant applications in cancer genomics as well. In this special issue of the Journal of Human Genetics, we aim to introduce various possibilities to our readers by having the developers of AI methods explain the diverse applications.