## **Studying the connectome at a large scale** ## **Yihe Weng, MSc; Rory Boyle PhD**, Trinity College Dublin **Relevant work:** [Shen et al., 2017][1], [Finn et al., 2015][2], [Greene et al., 2018][3] **Softwares/programs requirements:** [MATLAB][4] (Parallel Computing Toolbox), this is not free software but many institutions have a site licence. The code used in this presentation/chapter is freely available [here][5] and you will need 'Neuromethods_prepare_csv_CPM.m' and 'Neuromethods_main_analysis_CPM.m' for the demo. The Bioimage suite is [here][6] **AOMICS dataset:** [ID1000][7], slightly modified by Emin Serin [here][8]. **Modalities:** fMRI **Abstract:** Connectome-based predictive modelling (CPM) is a data-driven approach which enables the prediction of behavioural and cognitive phenotypes from functional connectivity data. CPM has been applied to successfully predict individual differences in various cognitive and behavioural phenotypes. A previous protocol paper outlined the method (Shen et al., 2017). Here, we present a flexible CPM method that is optimised for large neuroimaging datasets via the use of parallel computing. This approach enables researchers to account for possible site and scanner-related heterogeneity in multi-site neuroimaging datasets by controlling for site and/or scanner type as a covariate or by using leave-site-out cross-validation. Dr Rory Boyle will co-present this session. [1]: https://doi.org/10.1038/nprot.2016.178 [2]: https://doi.org/10.1038/nn.4135 [3]: https://doi.org/10.1038/s41467-018-04920-3 [4]: https://uk.mathworks.com/products/matlab.html [5]: https://github.com/rorytboyle/flexible_cpm [6]: https://bioimagesuiteweb.github.io/webapp/connviewer.html?species=human [7]: https://openneuro.org/datasets/ds003097 [8]: https://osf.io/5c2vr/files/