Deep learning methods for identifying nucleitide modifications from third generation sequencing datasets
Innovations in epitranscriptomics have resulted in the identification of more than 160 RNA modifications to date. These developments, together with the recent discovery of writers, readers, and erasers of modifications occurring across a wide range of RNAs and tissue types, have led to a surge in integrative approaches for transcriptome-wide mapping of modifications and protein–RNA interaction profiles of epitranscriptome players. RNA modification maps and crosstalk between them have begun to elucidate the role of modifications as signaling switches, entertaining the notion of an epitranscriptomic code as a driver of the post-transcriptional fate of RNA. Emerging single-molecule sequencing technologies and development of antibodies specific to various RNA modifications could enable charting of transcript-specific epitranscriptomic marks across cell types and their alterations in disease. My lab is interested in developing computational methods to predict the RNA modifications from third generation sequencing technologies and hence GPU resources are being requested to facilitate developing such algorithms.
Use of FutureSystems
We would like to see that tensorflow, pytorch and other popular deep learning libraries are installed on the system. However, if the systems are VMs we should be able to install them ourselves.
Scale of Use
I want to run a set of comparisons on 2 K80s or P100s and for each such comparison, I'll need about 5 days in a month for the next 5 months.