By Brandon Levy
Monday, December 17, 2018
Science is a process of trial and error. Most successful research publications are preceded by at least a few false starts and perhaps weeks or even months of tinkering to get experiments to work. For IRP senior investigator Carson Chow, Ph.D., this process of testing and throwing out one potential solution after another is an essential part of his research, so much so that he may go through thousands of iterations before arriving at one that works. However, rather than test each approach himself, he leverages the IRP’s considerable computing power to considerably accelerate the process of sorting the wheat from the chaff.
By Brandon Levy
Tuesday, November 27, 2018
The human genome comprises roughly three billion base pairs and around 20,000 protein-coding genes, according to recent estimates. That’s a lot of information crammed into the tiny nucleus of a cell, and it doesn’t even include the many genes that do not produce a protein or the fact that most genes come in multiple flavors that vary in different individuals. Add to that the phenomenon of an identical gene being either more or less active in two different people and you can quickly end up with genomic datasets that would overload nearly any computer. Fortunately for IRP senior investigator Daniel Levy, M.D., the NIH IRP has one of the few computer systems in the world that can handle this mountain of information.
By Brandon Levy
Thursday, August 2, 2018
For most of their history, computers have been limited to mindlessly executing the instructions their programmers give them. However, recent advances have given rise to the intertwined fields of artificial intelligence (AI) and machine learning, which focus on the creation of computer programs that can operate independently and even teach themselves to perform specific, specialized tasks. In 2013, the online PubMed database listed only 200 research publications related to ‘deep learning,’ a new type of machine learning that has shown success for particularly difficult tasks like object and speech recognition. Just four years later, in 2017, that number exceeded 1,100.
By Andy Baxevanis
Wednesday, June 13, 2018
Access to robust computing resources provides a critical foundation for advancing the wide variety of biomedical research taking place within the NIH’s Intramural Research Program (IRP). Whether performing molecular modeling simulations, generating whole-genome sequencing data, deducing the structures of biomolecules, or advancing drug discovery efforts, our ability to analyze large-scale biological and biomedical data strongly depends on our ability to employ computationally intensive approaches that produce interpretable results and advance translational efforts aimed at improving human health.
By Brandon Levy
Monday, February 26, 2018
For over a decade, my family shared our home with a short, fat beagle named Kayla Sue. She had big floppy ears, a tail as straight as an exclamation point, and a coat of fur that was a patchwork of white, brown, and black splotches. Her love of chasing small animals was matched only by her enthusiasm for eating, napping, and belly rubs. One of my best friends growing up, on the other hand, had a mean-spirited Dachshund named Rocky who would not let anyone outside his family touch his long, brown, sausage-shaped body. Meanwhile, one of my brother’s close childhood friends had two humongous, overly-friendly, black-and-brown German shepherds that would immediately bowl you over when you walked through the front door.
It doesn’t take a particularly sharp observer to notice that, despite being the same species, the more than 300 breeds of dog have remarkably different physical and behavioral traits. But what remains less clear even today are the specific biological roots that produce these widely varying attributes. And, perhaps more importantly, scientists seek to understand how learning about that immense diversity might help us improve the health of our canine companions – and ourselves.
By Hoo-Chang Shin
Tuesday, November 1, 2016
If you were going to train an artificial intelligence (AI) system to understand and accurately diagnose medical images, what kind of information do you think would be most effective: lots of general image data, or small amounts of specific data?
