Genomics in 3D: Overnight comparative studies mono-nucleosomes.

Thomas Bishop, Louisiana Tech University

Usage Details

This project will be to port and enhance the workflows that we have developed for the on-demand study of mono-nucleosomes to Blue Waters. The workflow is for high-performance high-throughput all-atom molecular dynamics. The workflow includes VMD tools for building a nucleosome with any sequence of DNA, amber tools for parameterization, NAMD for simulation and VMD for post-processing and visualization. The workflow integrates these tasks with job management on local or remote supercomputers and data sharing via iBIOMES.

Our previous efforts have demonstrated that for simulations containing approximately 250,000 atoms, 200 core is the limit of efficient parallel scaling for a single simulation. However, with modest supercomputing resources (100 to 200 nodes) we can sample ensembles consisting of 10 to 20 realizations of the nucleosome simultaneously. This allows us to generate hundreds of nanoseconds and potentially microseconds of sampling as an overnight run. The sampling may be an ensemble consisting of multiple replicas of the same system or the ensemble may contain any of the possible variations of the nucleosome arising from variations in DNA sequence.

Our scientific objective is to conduct a comparative study of mouse and human variants of the DNA sequence known as 54A54. This sequence contains the DNA sequence of Nucleosome A from the mouse mammary tumor virus (MMTV) surrounded by a flanking sequence of 54 basepairs on each side. Nucleosome A of the MMTV contains a hormone response element that has been exploited in a number of experimentally engineered gene constructs to create cells with a steroid dependent on-off switch for a specific gene. If steroids are added to the cell media, then the gene regulated by the MMTV construct is expressed. Without the addition of steroids the cell does not express the gene. A human analogue of the MMTV, known as HMTV, has been identified. It is associated with breast cancer and has been patented for use as a diagnostic tool for genetic screening. Specific mutations to the histone proteins that make up the nucleosome complex have been shown to alter the mobility of nucleosome A of the MMTV. To investigate the underlying cause of the altered mobility of nucleosome A we will shorten the 54A54 sequence to 12A12 and construct molecular models of the 21 possible positions of a nucleosome on this sequence. The positions include nucleosome A itself and 10 upstream and 10 downstream positions. A 20 basepair window correponds to two turns of the DNA helix (10bp/turn). Our sampling thus allows us to investigate the energy landscape associated with nucleosome positioning in the local neighborhood of nucleosome A.

The two students working on this project have only limited supercompuing experience. The goal of their internship is to rapidly bring them up to date and then have them continue their studies as PhD students. The students will benchmark and optimize runs of a single mono-nucleosome simulation on Blue Waters and will devise and validate a simulation strategy for sampling the 21 positions for human and mouse variants of the nucleosome associated with the 12A12 sequence. Specifically they will determine wither simple bundling strategies in which all 21 systems are run simultaneously or enhanced sampling methods developed in NAMD are the most “cost” efficient means of sampling our ensembles. The students will seek to improve the robustness, automation and efficiency of every aspect of the workflow, namely integrating modeling, simulation, data management, analysis and presentation of results.