Add training splits, clustering and supplementary data

initial commit