Researchers from Tokyo Metropolitan University have developed a suite of algorithms to automate the measurement of sister chromatid exchanges (SCE) in chromosomes under the microscope. Conventional analysis requires trained personnel and time, with variability between different individuals. The team’s machine learning-based algorithm boasts 84% accuracy and provides a more objective metric. This could be a game-changer for diagnosing disorders linked to an abnormal number of SCEs, such as Bloom’s syndrome.
DNA, the blueprint of life for all living organisms, is packaged inside complex structures called chromosomes. When DNA is copied, two identical strands known as sister chromatids are formed, each carrying exactly the same genetic information. In contrast to meiosis, sister chromatids do not need to undergo recombination during mitosis and are in most cases transmitted intact to daughter cells. However, when some form of DNA damage occurs, the body attempts to repair the damage using the remaining intact DNA as a template. During this repair process, it often happens that certain parts of the sister chromatids are exchanged with each other. During this repair process, it often happens that certain parts of the sister chromatids are exchanged with each other. This “sister color exchange” (SCE) is not harmful in itself, but too much can be a good indicator of some serious disorders. Examples include Bloom syndrome: affected individuals may be predisposed to cancer.
To measure SCEs, common methods involve experienced clinicians looking at stained chromosomes under a microscope, trying to identify the tell-tale “swapped” sections of sister chromatids. Not only is this work intensive and slow, but it can also be subjective, depending on how the human eye perceives the features. A fully automated analysis of microscope images would save time and give objective measurements of the number of SCEs, for more consistent diagnoses in different clinical settings.
Now, a team led by Professors Kiyoshi Nishikawa and Kan Okubo from Tokyo Metropolitan University have developed a series of algorithms that use machine learning to count SCEs in images. They combined separate methods, one to identify individual chromosomes, another to tell if SCEs are present, and finally another to group and count them, giving an objective, fully automated count of the number of SCEs in a microscope image. They found an accuracy of 84.1%, a level that is sufficient for practical applications. To see how it performed with real data, they collected images of chromosomes from artificially knocked-out cells BLM gene, the type of suppression seen in patients with Bloom’s syndrome. The team’s algorithm was able to provide measurements for SCE that were consistent with those given by human counters.
Work is currently underway to use the vast amounts of clinical data available to train the algorithm, with more improvements to come. The team believes that replacing manual measurement with full automation will help enable faster, more objective clinical analysis than ever before, and that this is just the beginning of what AI can bring to medical research.
This work was supported by JSPS KAKENHI Grant Numbers 22H05072, 25K09513 and 22K12170.
