Fig. 2
From: Reproducible biomedical benchmarking in the cloud: lessons from crowd-sourced data challenges
Challenge features. Challenges used cloud computing services for running and evaluating models including Google Cloud Platform, Openstack, Amazon Web Services, and IBM Cloud. Models were designed to run using either CPUs or GPUs. The type of data used in running and evaluation of models was either real data (obtained from patients or cell lines) or simulated using a computer algorithm. Challenges used genomic data, such as DNA sequencing, RNA sequencing, and gene expression; clinical phenotypes; and/or images. Models could be submitted to a challenge in the form of a galaxy workflow, docker image, or CWL (Common Workflow Language) workflow