Exploring Machine Learning for Rare Diseases: A Study Using Nature Methods

Recent advancements in high-throughput profiling methods, such as genomics and imaging, have revolutionized medical research and enabled an elaborate understanding of the molecular mechanisms and pathways underlying several diseases. Machine learning (ML) is a powerful tool to extract significant patterns from these high dimensional datasets. But due to its complexity, it usually requires a large sample size to identify essential patterns. However, when dealing with rare diseases, the clinical cases tend to be quite limited, making it a difficult task to perform ML on these samples. In this Perspective, we focus on addressing the specific challenges related to employing ML for rare disease research, considering the limited samples. We also emphasize potential solutions to the problem and suggest that ML techniques with limited samples should be given preference and be developed further. With the goal of further advancing rare diseases, this could also potentially broaden ML applications to cases which have few samples with high-dimensional data.