Include ENCODE datasets from both normal and cancer cell lines
Mining cancer data in the cloud is great, but to enable ongoing research there should be a connection to specimens so researchers can pursue followup studies. This will require storing data about specimens from studies such as TCGA - where they are, how they can be accessed and what consent they are governed by. Just as the data from publications should be made available to allow reproduction of results, so should samples ...more »
GPU technologies are rapidly becoming useful for speeding up some workflows by orders of magnitude. It would be useful to have some GPU resources available for cloud computing.
Datasets containing the quantitative inventory of proteins in TCGA tumors are beginning to become available. Both mass spectrometry and affinity-based technologies are generating these data. The cloud should provide a means to connect these data to corresponding TCGA data.
Provide a series of online short videos and short courses that will help users adopt the new tools and instructors to incorporate into courses. (Maybe this is obvious, but high-quality tutorials and case studies take significant time to develop.)
A sample could be analyzed for DNA sequence variations, structural variations, CNVs, Gene or transcript isoform expression, genome-wide methylation patterns, ChIP-seq for specific transcription factors, metabolomic or proteomic analysis, and other molecular profiles. A framework that allows a researcher to readily identify all molecular data types associated with a particular sample and integrate the results of such analyses ...more »
To stimulate learning as much as possible, as quickly as possible, the data cloud could have a utility where interested parties could pose "crowd-sourcing" challenges, e,g, Kaggle. Indeed, Harold Varmus, NCI & leaders in cancer & genomics could pose the leading questions they would like bright people to take a run at answering, e.g. Hilbert's 23 problems
Track data provenance and permissions, including IRB approvals and patient consent and be able to support different levels of permissions rather than insisting on uniform consent
Provide access to BAM files for TCGA miRNA sequencing data for analysis
Realign sequencing data sets to a common genome version
Provide access to level 1 and 2 data for TCGA copy number array data for analysis
Analyze exomic sequence of paired tumor and normal samples, including variant calls
Provide access to level 1 and 2 data for TCGA DNA methylation data for analysis
For those of us who don't want to use the cloud workflow, please make it so we still have access to all the raw data. Please don't lock us into your analytic approaches!
Most current approaches for BigData analysis involve moving data to a server, HPC infrastructure or cloud where the software tools and reference databases are pre-configured. This is inefficient since this approach requires making redundant copies of data each time and additional costs/time associated with moving data back and forth. Since there is no single tool or workflow to analyze genomic data, multiple copies ...more »