Scoping review and classification of deep learning in medical genetics

Abstract

Deep learning (DL) is applied in many biomedical areas. We performed a scoping review on DL in medical genetics. We first assessed 14,002 articles, of which 133 involved DL in medical genetics. DL in medical genetics increased rapidly during the studied period. In medical genetics, DL has largely been applied to small data sets of affected individuals (mean = 95, median = 29) with genetic conditions (71 different genetic conditions were studied; 24 articles studied multiple conditions). A variety of data types have been used in medical genetics, including radiologic (20%), ophthalmologic (14%), microscopy (8%), and text-based data (4%); the most common data type was patient facial photographs (46%). DL authors and research subjects overrepresent certain geographic areas (United States, Asia, and Europe). Convolutional neural networks (89%) were the most common method. Results were compared with human performance in 31% of studies. In total, 51% of articles provided data access; 16% released source code. To further explore DL in genomics, we conducted an additional analysis, the results of which highlight future opportunities for DL in medical genetics. Finally, we expect DL applications to increase in the future. To aid data curation, we evaluated a DL, random forest, and rule-based classifier at categorizing article abstracts.

Keywords: Artificial intelligence; Deep learning; Machine learning; Medical genetics; Medical genomics.

Figure 1. PRISMA schema for data collection and categorization.

Although we performed a scoping review, this schema was adapted from the one used for systematic reviews and is used with appropriate permission and citation as described in the guidelines.^, Sources used include Clinical Genomic Database (CGD), Face2Gene, OMIM, and PubMed.^- PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

Figure 2. Articles per year and characteristics of studied individuals.

A. Number of articles per year binned as category 1 (articles on deep learning [DL] applied to genetic conditions). Articles from 2021 includes observed articles as well as projected articles, the latter was calculated on the basis of the observed trend during the depicted time period (January 2015-June 2021). B. Distribution of genetic conditions studied using DL. C. Number of individuals with the studied genetic conditions included in each study. Further details are available in Supplemental Table 3.

Figure 3. Characteristics of methods used.

A. Types of clinical input data analyzed via deep learning (DL). B. Types of DL methods used in each article. C. Categorization of the primary use of DL in each article. Further details are available in Supplemental Table 3. BERT, bidirectional encoder representations from transformers; CNN, convolutional neural network; ECG, electrocardiogram; EEG, electroencephalogram; RNN, recurrent neural network.

Figure 4. Geographic distribution of articles.

A. Location of the corresponding author(s) for each of the 134 articles. B. Location of study populations for articles with available data.