### INTRO ### This Zenodo file collection contains transcriptome prediction models built for PrediXcan, as well as concatenated raw results from SPrediXcan. Article: Multivariate adaptive shrinkage improves cross-population transcriptome prediction for transcriptome-wide association studies in underrepresented populations. Authors: Daniel S. Araujo, Chris Nguyen, Xiaowei Hu, Anna V. Mikhaylova, Tim Thornton, Chris Gignoux, Leslie Lange, François Aguet, Kristin Ardlie, Kent D. Taylor, Peter Durda, Craig Johnson, Elaine Cornell, Xiuqing Guo, Yongmei Liu, Russ Tracy, George Papanicolaou, Michael H. Cho, Stephen S. Rich, Jerome I. Rotter, NHLBI TOPMed Consortium, Hae Kyung Im, Ani Manichaikul, Heather E. Wheeler. ### TRANSCRIPTOME PREDICTION MODELS ### There are transcriptome prediction models built using three different methods (Elastic Net [EN], MatrixeQTL, Multivariate adaptive shrinkage [MASHR]), for three different tissues (Monocytes [Mono], Peripheral Blood Mononuclear Cells [PBMC], T-cells) and up to four populations (African Americans [AFA], Chinese [CHN], European [EUR], Hispanic/Latino [HIS]). For PrediXcan, only the ".db" files are necessary. For SPrediXcan, however, it's also necessary to use the respective covariance file. For more detailed instructions on how to use those files, please refer to the MetaXcan GitHub repo (https://github.com/hakyimlab/MetaXcan). ### SPREDIXCAN RESULTS ### Raw, concatenated SPrediXcan results using our models and GWAS summary statistics from PAGE and PanUKBB are also provided. For a better description of what each column in the SPrediXcan output means, please refer to the Wiki page of the MetaXcan GitHub repo (https://github.com/hakyimlab/MetaXcan/wiki).