Overlapping but Asymmetrical Relationships Between Schizophrenia and Autism Revealed by Brain Connectivity
Yujiro Yoshihara,Giuseppe Lisi,Noriaki Yahata,Junya Fujino,Yukiko Matsumoto,Jun Miyata,Gen-ichi Sugihara,Shin-ichi Urayama,Manabu Kubota,Masahiro Yamashita,Ryuichiro Hashimoto,Naho Ichikawa,Weipke Cahn,Neeltje E. M. van Haren,Susumu Mori,Yasumasa Okamoto,Kiyoto Kasai,Nobumasa Kato,Hiroshi Imamizu,Rene S. Kahn,Akira Sawa,Mitsuo Kawato,Toshiya Murai,Jun Morimoto,Hidehiko Takahashi
DOI: https://doi.org/10.1101/403212
2020-01-01
Schizophrenia Bulletin
Abstract:Although the relationship between schizophrenia spectrum disorder (SSD) and autism spectrum disorder (ASD) has long been debated, it has not yet been fully elucidated. To address this issue, we took advantage of dual (ASD and SSD) classifiers that discriminate patients from their controls based on resting state brain functional connectivity. An SSD classifier using sophisticated machine-learning algorithms that automatically selected SSD- specific functional connections was applied to Japanese datasets including adult patients with SSD in a chronic stage. We demonstrated good performance of the SSD classification for independent validation cohorts. The generalizability was tested by USA and European cohorts in a chronic stage, and one USA cohort including first episode schizophrenia. The specificity was tested by two adult Japanese cohorts of ASD and major depressive disorder, and one European cohort of attention-deficit hyperactivity disorder. The weighted linear summation of the classifier’s functional connections constituted the biological dimensions representing neural liability to the disorders. Our previously developed robust ASD classifier constituted the ASD dimension. Distributions of individuals with SSD, ASD and healthy controls were examined on the SSD and ASD biological dimensions. The SSD and ASD populations exhibited overlapping but asymmetrical patterns on the two biological dimensions. That is, the SSD population showed increased liability on the ASD dimension, but not vice versa. Furthermore, the two dimensions were correlated within the ASD population but not the SSD population. Using the two biological dimensions based on resting-state functional connectivity enabled us to quantify and visualize the relationships between SSD and ASD.