How to Unlock the Secrets of DNA With High-Throughput Computing

The binding power of DNA is well known – bases, bound into complementary pairs, are the building blocks of its double helix structure. Short sequences of DNA (15 to 40 base pairs in length) also chemically bond to small nanoparticles.

Unpaired nucleotides give DNA its Velcro-like quality. However, scientists are unsure how short strands of DNA actually find their complement and bind. Many researchers have posed theories for DNA sequences longer than 100 bp, but they have yet to develop theories for shorter sequences.

Dan Hinckley, a graduate student in the research group of Juan J. de Pablo at the University of Wisconsin–Madison (UWM), US, is developing an improved coarse-grained model of DNA. This model is designed to enable the study of rare events that are intractable with all-atom simulations – specifically, how short strands of complementary and partially complementary DNA come together.

“These coarse-grained DNA models are representations of biomolecules one level above an all-atom representation. We combine groups of atoms together into a single site,” says Hinckley. This coarse model is the key to uncovering the finer details he is seeking: “We need a model that allows us to simulate sufficiently long sequences for long enough times to see rare events, where the single-stranded DNAs come together to form a duplex. We also need to generate ensembles of trajectories, so we can be confident in our statistics.”

“This is where high-throughput computing (HTC) and the Center for High Throughput Computing at UWM come in. Because coarse-grained models are computationally inexpensive and each simulation is performed independently, we’re able to take advantage of resources like HTCondor and Open Science Grid to run large batches of calculations. And, because we're running large batches, we can be confident that the results we obtain are representative of all possible trajectories or mechanisms by which the DNA strands will come together to form double-stranded DNA,” explains Hinckley.

Coarse-grained modeling is still very much an art form: researchers have yet to discover a systematic procedure that yields coarse-grained models applicable to a wide range of systems. “Coarse-grained modeling presents a great challenge. Defining the interactions and adjusting them such that we capture all of the physics relevant to the process takes a lot of simulation power," Hinckley says. “It is necessary to establish rigorously that the model reproduces experiments and other measured phenomena. To do it right, it is critical that we have access to HTC resources.” -- Amber Harmon, © i SGTW