2025 年 42 巻 4 号 p. 574-579
Ultarhigh–dose methylcobalamin was approved as a treatment for amyotrophic lateral sclerosis (ALS) in 2024 in Japan. Although ultrahigh–dose methylcobalamin had not shown efficacy in a Phase 2/3 trial targeting patients with onset within 3 years, a Phase 3 trial targeting patients with onset within 1 year showed that ultrahigh–dose methylcobalamin significantly suppressed the decline in ALSFRS–R at 16 weeks of administration compared to placebo. The results indicate that early treatment is important for ALS. In 2025, tofersen, a nucleic acid drug for ALS caused by SOD1 gene mutations, was also launched in Japan. Tofersen is also undergoing a Phase 3 ATLAS trial targeting people with SOD1 mutations who have not yet developed ALS and whose plasma neurofilament light (NfL) concentrations are below the standard. The traial design is based on presymptomatic familial ALS studies showing that increases in serum NfL precede clinical onset by 6 to 12 months. In other words, NfL can be a predictive biomarker for the onset of familial ALS. On the other hand, abnormality in NfL is not specific to ALS, and its application to early or presymptomatic stages of sporadic ALS would be limited. We hope to use muscle ultrasonography and induced pluripotent stem cells (iPSCs) for early and pre–symptomatic diagnosis of sporadic ALS. In ALS, fasciculation occurs as a functional abnormality at the macro level before muscle weakness or atrophy occurs. The use of muscle ultrasonography, which is excellent at detecting fasciculation, is useful for early diagnosis after the onset of ALS. We worked on predicting whether a patient has ALS or not by using deep learning with images of motor neurons differentiated from iPSCs of ALS patients. We trained artificial intelligence (AI) using image data of motor neurons from iPSCs of healthy volunteers and ALS patients, and constructed an AI model that can distinguish between healthy volunteers and ALS motor neurons. It was revealed that this model can classify healthy volunteers and ALS patients with high accuracy by focusing on the cell body and neurites of motor neurons. Although further verification using more information is required, it is hoped that the combination of AI and iPSC technology will contribute to the prediction of the onset and pre–symptomatic diagnosis of ALS.
