research
What we did and how they interact with each other
Autism
Oxytocin series
Until 2014, the therapeutic effects of oxytocin on autism were not experimentally validated. By conducting a case-control study (Watanabe et al., 2012, PLoS One) and a randomised, double-blind clinical trial (Watanabe et al., 2014, JAMA Psychiatry), we revealed macroscopic neurobiological mechanisms by which oxytocin mitigates autistic social behaviours, and provided one of the first behavioural and neural evidence for the beneficial effects of the neuropeptide.
In addition, we showed that this effect is attributable to elevation of a neurochemical substance in a focal brain region (Aoki, Watanabe et al., 2015, Mol. Psychiatry), and demonstrated reproducibility of these findings in a follow-up study (Aoki et al., 2015, Brain).
In a Brain paper in 2015, we reported that such beneficial effects can be seen not only in experimental settings but also in clinical examinations after six-week administration (Watanabe et al, 2015, Brain). We also have developed a novel deep-learning-type algorithm and identified genetic variations that enable us to predict the efficacy of oxytocin in each autistic individual prior to actual administration (Watanabe et al., 2016, SCAN).
Brain architecture
We also found new neuroanatomical features in autism. In 2015, we reported atypically under-development of rich-club organisation in autism (Watanabe and Rees, 2015 Sci Rep). In 2016, we found imbalance of relative grey matter volume in autistic individuals (Watanabe and Rees, 2016 Sci Rep).
In 2017, we identified atypical brain dynamics in autism and revealed that atypically stable neural dynamics underpin not only core symptoms of the prevalent neurodevelopmental disorder but also its unique cognitive styles (Watanabe and Rees, 2017 Nat Comms).
Energy landscape analysis
What's the advantage of energy landscape analysis?
In human neuroimaging studies, conventional analysis methods —such as a univariate regression analysis and multivariate pattern analysis— are effective in identifying a single or limited pattern of brain activity that is time-locked to a discrete psychological event. However, they are generally poor at characterising dynamic changes in neural activity over time that are not time-locked to explicit cognitive events and are implicitly occurring before/after apparent behavioural responses.
In theoretical neuroscience studies, most of the mathematical investigations about neural dynamics do not aim to provide practical analysis tools, and thus, it has been difficult to exploit their results for actual analyses of empirically-obtained multiple neural train data.
Consequently, it was difficult to capture neurobiological principles of dynamic brain activity in fluctuating neural data that should support some critical human cognitive phenomena.
In contrast, energy landscape analysis, which has been a favourite method in molecular dynamics studies, can characterise such spatiotemporally high dimensional, fluctuating neural activity data. This analysis enables us to automatically identify dominant brain activity states in fluctuating neural train data, and depict whole-brain neural dynamics as staying in and transitions between such major dominant brain states.
What we have done with this.
In 2013, we shown a possibility that the energy landscape model can be used for functional MRI signals (Watanabe et al., 2013 Nat Comms), and in 2014, actually identified brain dynamics during unstable visual perception by this data-driven approach (Watanabe et al., 2014 Nat Comms). In 2017, this method enabled us to reveal intrinsic neural dynamics that underpin both symptoms and cognitive styles in autism (Watanabe and Rees, 2017 Nat Comms). We also found age-related changes in brain dynamics with this analysis (Ezaki et al., 2018 Human Brain Mapp). A review paper in 2017 summarises the method (Ezaki et al., 2017 Phil. Trans. R. Soc. A.).