## **Connectome-based Machine Learning using NBS-Predict** ## **Emin Serin, MSc**, Charite Berlin **Relevant work:** Serin et al. (2021) [NBS paper][1] in NeuroImage and the book chapter [preprint][2] **Softwares/programs requirements:** [MATLAB][3] (R2016b or newer) including Statistics and Machine Learning Toolbox, Parallel Computing Toolbox (optional) as well as [NBS-Predict][4] **AOMICS dataset:** [ID1000][5] with some modifications (see 'files' section below for data used in this presentation). **Modalities:** fMRI **Abstract:** In this tutorial, I will demonstrate NBS-Predict, a prediction-based extension of the Network-based Statistic (Zalesky et al., 2010). NBS-Predict aims to alleviate the curse of dimensionality, lack of interpretability, and problem of generalizability. By combining the powerful features of machine learning and graph theory in a crossvalidation structure, it provides a fast and convenient tool to identify neuroimaging-based biomarkers with high generalizability. Unlike generic machine leaming algorithms, results derived from the toolbox are straightforwardly interpretable. NBS-Predict comes with a user-friendly graphical user interface (GUI) developed on MATLAB. Thu it does not require any programming expertise. The toolbox provides an interactive viewer to visualize the results. The extensive user manual and usage tutorials are included in the toolbox. [1]: https://doi.org/10.1016/j.neuroimage.2021.118625 [2]: https://osf.io/4ghpm/ [3]: https://uk.mathworks.com/products/matlab.html [4]: https://github.com/eminSerin/NBS-Predict [5]: https://openneuro.org/datasets/ds003097