Content with tag biology .

Scientists rely on XSEDE to compare massive amounts of genetic data across species

By: Faith Singer-Villalobos, TACC

In many non-monogamous species, females provide all or most of the offspring care. In monogamous species, parental care is often shared. In these frogs, parental care includes transporting tadpoles one by one after hatching to small pools of water. In the non-monogamous strawberry poison frog (Oophaga pumilio, left) moms perform this task; however, in the monogamous mimic poison frog (Ranitomeya imitator, right) this is dad's job. Credit: Yusan Yan and James Tumulty

Why are some animals committed to their mates and others are not?

According to a new study led by researchers at The University of Texas at Austin that looked at 10 species of vertebrates, evolution used a kind of universal formula for turning non-monogamous species into monogamous species — turning up the activity of some genes and turning down others in the brain.

"Our study spans 450 million years of evolution, which is how long ago all these species shared a common ancestor," said Rebecca Young, research associate in UT Austin's Department of Integrative Biology and first author of the study published this week in the journal Proceedings of the National Academy of Sciences.

The authors define monogamy in animals as forming a pair bond with one mate for at least one mating season, sharing at least some of the work of raising offspring and defending young together from predators and other hazards. Researchers still consider animals monogamous if they occasionally mate with another.

The researchers studied five pairs of closely related species – four mammals, two birds, two frogs and two fish — each with one monogamous and one non-monogamous member. These five pairs represent five times in the evolution of vertebrates that monogamy independently arose, such as when the non-monogamous meadow voles and their close relatives the monogamous prairie voles diverged into two separate species.

The researchers compared gene expression in male brains of all 10 species to determine what changes occurred in each of the evolutionary transitions linked to the closely related animals. Despite the complexity of monogamy as a behavior, they found that the same changes in gene expression occurred each time. The finding suggests a level of order in how complex social behaviors come about through the way that genes are expressed in the brain.

This study covers a broader span of evolutionary time than had been explored previously. Other studies have looked at genetic differences related to evolutionary transitions to new traits, but they typically focus on animals separated by, at most, tens of millions of years of evolution, as opposed to the hundreds of millions of years examined with this study.

"Most people wouldn't expect that across 450 million years, transitions to such complex behaviors would happen the same way every time," Young said.

Researchers examined gene activity across the genomes of the 10 species, using RNA-sequencing technology and tissue samples from three individuals of each species. The scientists detected gene-activity patterns across species using bioinformatics software and Wrangler. Operated by the Texas Advanced Computing Center (TACC), Wrangler is a data-intensive supercomputer funded by the National Science Foundation that is part of the XSEDE ecosystem.

Using a software package, OrthoMCL, the team was able to arrange genes from distantly related species — such as a fish and a mammal — into groups based on sequence similarities. This allowed them to identify the common evolutionary formula that led to pair bonds and co-parenting in the five species that behave monogamously.

At least five times during the past 450 million years, evolution used a kind of universal formula for turning animals monogamous — turning up the activity of some genes (red) and turning down others (blue) in the brain. Researchers studied five pairs of closely related species – four mammals, two birds, two frogs and two fish — each with one monogamous and one non-monogamous member. They found 24 genes with similar expression patterns in monogamous males. Illustration credit: The University of Texas at Austin.

"Wrangler is set up with a relational database that allows individual computational steps to go back and talk to this database, and pull up the information it needs without any timeout errors," Young said. "We've been able to run all of our species together on Wrangler using OrthoMCL, and at this point we haven't even maxed out what Wrangler is capable of doing."

According to Young, with traditional online databases she was only able to identify about 350 comparable genes across these 10 species; however, when she ran OrthoMCL on Wrangler, she identified almost 2,000 genes that are comparable across all of the species.

"This an enormous improvement from what is available," Young said. "When you add this up across 10 species, you have an enormous amount of data. We're starting at a minimum of 80,000 genes that we're going to compare in all pairwise combinations for over three billion comparisons to perform and organize in total. The Wrangler supercomputer helped make this science possible."

The paper's other UT Austin authors are senior author professor Hans Hofmann and professor Steven Phelps.

Authors at other institutions are Michael Ferkin (University of Memphis), Nina Ockendon-Powell (University of Bristol), Veronica Orr (University of California, Davis), Ákos Pogány (Eötvös Loránd University), Corinne Richards-Zawacki (University of Pittsburgh), Kyle Summers (East Carolina University), Tamás Székely (University of Bath), Brian Trainor (University of California, Davis), Araxi Urrutia (University of Bath and Universidad Nacional Autónoma de México), Gergely Zachar (Semmelweis University) and Lauren O'Connell, a former UT Austin graduate student (Stanford University).

This work was supported by the Alfred P. Sloan Foundation, the National Science Foundation, the National Institutes of Health, and the Hungarian Scientific Research Fund.

XSEDE-allocated resources at TACC and PSC, together with other national resources, model inositol phosphate interactions with HIV-1 structural proteins


Published on August 9, 2018 by Jorge Salazar

The naturally-occurring compound IP6 (red) facilitates the formation and assembly of HIV-1 structural proteins, results from XSEDE Stampede2 and Anton2 systems show. Image courtesy of Perilla et al.

HIV-1 replicates in ninja-like ways. The human immunodeficiency virus slips through the membrane of vital white blood cells. Inside, HIV-1 copies its genes and scavenges parts to build a protective bubble for its copies. Scientists don't understand many of the details of how HIV-1 can fool our immune system cells so effectively. The virus infects 1.2 million people in the U.S. and 37 million people worldwide in 2018. The Extreme Science and Engineering Discovery Environment (XSEDE) awarded supercomputer time that helped model a key building block in the HIV-1 capsid, its protective shell, and its interaction with a family of small molecules critical for viral function. The discovery could lead to novel strategies for potential therapeutic intervention in HIV-1 replication.


Juan Perilla, Department of Chemistry and Biochemistry, University of Delaware.
Scientists found that the naturally occurring compound inositol hexakisphosphate (IP6) promotes both assembly and maturation of HIV-1. "We discovered, in collaboration with other researchers, that HIV uses this small molecule to complete its function," said Juan R. Perilla, Department of Chemistry and Biochemistry, University of Delaware. "This is a molecule that's extremely available in human cells and in other mammalian cells. HIV has evolved to make use of these small molecules present in our cells to essentially be infectious." Perilla co-authored the study in the journal Nature in August 2018.


Perilla ran simulations of inositol phosphate interactions with HIV structural proteins CA-CTD-SP1 using the Nanoscale Molecular Dynamics (NAMD) software through allocations awarded by XSEDE, which is funded by the National Science Foundation (NSF). "XSEDE provides a unique framework which allows us to use computational resources that are tailored to the needs of a particular scientific problem. In addition, we benefit from the HPC training opportunities provided by XSEDE, which allows us to develop novel analysis tools," Perilla said.


Model. a, Diagram of HIV-1 Gag. Dotted lines indicate protease cleavage sites. b, Diagram of Gag organization in immature virions (left). Following cleavage of Gag by protease (that is, maturation), CA re-organizes to form a mature core around viral RNA (right). c, d, Surface representations of the CASP1 and CA hexamers in the immature (c) and mature virus (d), with IP6 shown in its binding sites. The marked rearrangement of CA upon maturation is evident, as is the change in IP6 binding site between immature and mature viruses. CA NTD , blue; CA CTD , orange; 6HB, purple; IP6 , red. Image courtesy of Perilla et al.
The Perilla group used the XSEDE-allocated systems Stampede2 at the Texas Advanced Computing Center and Bridges at the Pittsburgh Supercomputing Center (PSC), as well as other national resources, to run simulations of the Inositol phosphates IP3, IP4, IP5, IP6 and their interactions with HIV proteins CA-CTD-SP1. "What these systems allowed us to do is establish what the molecular interactions are between the HIV proteins and this small molecule. With them we were able to test the hypothesis that it was stabilizing a particular part of the protein using molecular dynamics. I think Stampede2 and Bridges are great machines, and it's extremely beneficial to the scientific community to have resources like these available on a merit-based system," Perilla said.
"What I would like the public to know is that it's important that these large-scale machines are available," he added. "They are not just a replacement of a small [campus] cluster. These machines really enable new science. If you didn't have machines of this scale, you couldn't do the kind of science that we do."


Side view of the six-helix bundle showing two rings of Lys290 and (cyan) with bound IP6 in the middle. The bundle which holds together the Gag hexamer and facilitates the formation of a curved immature hexagonal lattice underneath the viral membrane. Image courtesy of Perilla et al.
Perilla described the increasing use of the 'computational microscope,' the combination of supercomputers with laboratory data. "With the computational microscope, you can see how things move. Many experimental techniques are just a snapshot. With the computational microscope, you can actually see how things are moving," he said.


Supercomputer modeling of how building blocks of HIV-1 move in time made a difference in this study. "That discovery opens a door for development of new treatments. It's a therapeutic target. Because of that, it makes it very appealing for drug development and therapeutic development," Perilla said.

There remains much to be learned about how HIV-1 behaves, said Perilla. "We're basic scientists. NSF's mission is to understand these systems as living organisms. The overall idea is that we want to understand the virus as a biological problem and ultimately this knowledge will be used to derive therapeutics," Perilla said.

The study, "Inositol phosphates are assembly cofactors for HIV-1," was published in the journal Nature on August 1, 2018. The study authors are Robert A. Dick and Volker M. Vogt of Cornell University; Kaneil K. Zadrozny, Jonathan M. Wagner, Barbie K. Ganser-Pornillos, and Owen Pornillos of the University of Virginia; Chaoyi Xu and Juan R. Perilla of the University of Delaware; Florian K. M. Schur of the European Molecular Biology Laboratory and the Institute of Science and Technology Austria; Terri D. Lyddon, Marc C. Johnson, and Clifton L. Ricana of the University of Missouri. The National Institutes of Health funded the research. This work used the Extreme Science and Engineering Discovery Environment, which is supported by National Science Foundation grant number OCI-1053575. The work also relied on computation on the special-purpose Anton 2 system, which is made available without cost by D.E. Shaw Research and hosted by PSC with funding from grant R01-GM116961 from the National Institutes of Health.

Link to Cornell University press release:


Function Follows Form

Simulations on XSEDE Resource plus Lab Work on Frog Neuromuscular Junction Sheds Light on Human Diseases

When a nerve cell passes a message to its neighbors, it must do so via chemicals sent across the synapse—a small space between the cells. Early researchers studied a synapse called the frog neuromuscular junction (NMJ) because it is large and easy to work with. But its different organization and behavior compared to mammalian synapses led many scientists to dismiss it as not relevant to human biology. A team of University of Pittsburgh and XSEDE Extended Collaborative Support Service (ECSS) scientists performed simulations on the XSEDE-allocated Bridges supercomputer at the Pittsburgh Supercomputing Center (PSC) and parallel lab experiments on the frog and mouse NMJ. They showed that, when reorganized into the same geometric pattern as in the mouse, the components of the frog NMJ act like those in the mouse. Lessons from the work are already being used to design candidate drugs to treat human neuromuscular diseases.

Why It's Important:

The synapse—the small space between a nerve cell and its neighbor—is where the rubber meets the road in just about everything our brains, senses and muscles do. Whenever a nerve cell fires, whether to tell a muscle to twitch, convey the smell of strawberries, or trigger a hallucination in someone with schizophrenia, it must pass the message along with a burst of neurotransmitter chemicals released into the synapse. Whether we're trying to restore use of the legs to someone with spinal cord damage, treat a psychiatric condition, or protect nerve cells from a degenerative neuromuscular disease, synapses are likely to play an important role.

"It's worth saying that neurotransmitter release underlies everything in the brain. We still don't know how that works … and it's a basic process in the nervous system."—Anne Homan, University of Pittsburgh

But for all the synapse's importance, scientists still face challenges in understanding exactly how it works. The mouse neuromuscular junction (NMJ)—the synapse between a nerve cell that makes a mouse's muscle move and a muscle cell—is similar to many synapses in humans. It's also proven a good stand-in for scientists studying human neuromuscular diseases. Historically, some of the first studies of how cells communicate with one another in the nervous system began with the frog NMJ. That's because it's large enough to easily see in a microscope. It also allows variety of experimental manipulations. But because frogs are amphibians and not mammals, scientists have disagreed about the relevance of the frog NMJ to human biology.

"The frog neuromuscular junction is easy to get to and hundreds of microns in length. It's a simple model in which you have a monster synapse … with chemical release sites arranged in a regular pattern like ties on a railroad track. It's easy to manipulate, characterize and study."—Stephen Meriney, University of Pittsburgh

Both the organization of the frog and mouse NMJs and their behavior differ. But the chemical actors—proteins, neurotransmitters and other components—are identical in the frog and mouse NMJ. This fact led the team of Stephen Meriney of the University of Pittsburgh; his graduate student Anne Homan; Rozita Laghaei, an XSEDE ECSS expert at PSC; and colleagues at PSC and Pitt to use these systems as models to determine how structure influences function.

How XSEDE and PSC Helped:

Homan carried out a series of lab experiments on frog and mouse NMJs. Meanwhile, Laghaei used the MCell software to create a virtual nerve cell/muscle cell pair running on PSC's Bridges supercomputer. MCell was developed by the National Center for Multiscale Modeling of Biomedical Systems, made up of PSC, Pitt, the Salk Institute and Carnegie Mellon University. The task was challenging, since MCell must carry out complex calculations that depend on each other's results to proceed. So the common supercomputing strategy of splitting the problem into many small parts that can be performed independently at the same time wasn't possible. Laghaei had to run the same simulation tens of thousands of times, each differing by random changes. This approach built up a representative sample of the behavior of neurotransmitter release by the nerve cell, bit by bit. The method, called Monte Carlo simulation, required massive computer memory. Bridges' unique large-memory nodes made the computation manageable.

"The MCell part of the study was to model as close as possible what the experimentalists see … It involved lots of computations. We ran several simulations exactly the same except for sequences of random numbers that we use to model … transmitter release at neuromuscular junctions (which is a stochastic process) and did that experiment ten thousand times … It required huge memory; the output of the simulation kept track of every particle."—Rozita Laghaei, PSC

The comparison of the lab and computer experiments shed a lot of light on the scientists' questions. In the computer model, Laghaei reorganized the components of a frog NMJ so that they were laid out in the same geometric pattern as in a mouse NMJ. She found that the behavior of the new system was identical to what Homan saw in the real mouse. The difference between the two species lay entirely in how the NMJ components were organized, and not in a more fundamental difference. This result shows that the frog system is relevant to human synapses after all. Later simulations showed that the scientists could tune the behavior of the NMJ at will by changing the geometry of the components. The group reported their results in two papers in the Journal of Neurophysiology in November and December 2017. The work is a vital first step in understanding the NMJ and other synapses. Colleagues in Pitt's Department of Chemistry are now using the lessons learned to understand the fundamental changes that occur in diseases of synapses. They aim to design new drug candidates to treat neuromuscular diseases in humans.

View from the inside of a nerve cell. The synapse and the muscle cell are not pictured, but would be below the nerve cell's cell membrane, at the bottom. In the frog NMJ (A), neurotransmitter-containing packets (vesicles) waiting to be dumped into the synapse are arranged in two rows. (Vesicles are in red, calcium channels below the vesicles are small red dots, and the calcium ions diffusing in the nerve terminal are represented as small blue or yellow dots.) In the mouse (B), the vesicles are organized in clusters that each contain two vesicles. Simulations on Bridges showed that the frog system, when rearranged in clusters like the mouse, began to behave like the mouse NMJ.

Testing the Footing

XSEDE Resources Help Univ. of Chicago Team Simulate Cell Movement, Upending Scientific Expectations

Aug. 31, 2018

The movement of white blood cells to fight infections and the spread of cancer cells both rely on the same natural process. The cell reaches out to a new surface with a lamellipodium—a kind of tiny foot that tests the surface like we'd test ice before stepping onto it. As part of a multi-institutional collaboration, a team from the University of Chicago simulated how the lamellipodium works, using XSEDE resources and online training tools in concert with laboratory experiments. Their virtual cells duplicated their lab findings perfectly, showing how integrin and fibronectin—two proteins scientists had previously not expected to play a role—tug on the surface before the cell commits to moving onto it. The discovery points to possible ways for doctors to encourage good cell movement and discourage bad cell movement.

Why It's Important:

The ability of the cells in our body to move around lies at the heart of vital life functions. It allows white blood cells to move into tissues to fight infections. It helps organs in developing embryos organize and grow. More ominously, it also enables cancer cells to move and spread. Fully understanding cell motion could help doctors better encourage the good cell movements and stop the bad ones.

The vital first step in cell motion is when the cell forms a lamellipodium—meaning "thin sheet foot." With its lamellipodium, the cell reaches out to test a surface, called a substrate, before the cell moves onto it. The amount of extension of the lamellipodium depends on the rigidity of the substrate. This motion is similar to how we'd gently step onto ice to make sure it's solid before we put our weight onto it. If the substrate is solid enough, the cell will move onto it. If not, the cell doesn't venture onto the "ice." By encouraging, or discouraging, that interaction, we could speed or stop motion of a cell into a given place or tissue.

"Sensing by the cell determines the first phase of cell adhesion to a substrate. Other things can then happen, but the very initial contact between a cell and a substrate is dependent on the lamellipodium."—Tamara Bidone, University of Chicago

Gregory Voth of the University of Chicago and his postdoc Tamara Bidone wanted to understand how a lamellipodium tests a substrate's rigidity, and to clear up some disagreements among scientists over how it works. With Patrick Oakes at the University of Rochester, NY, and colleagues at Chicago, they turned to a mix of laboratory measurements and computer simulations of lamellipodia. The latter employed the XSEDE system Bridges at the Pittsburgh Supercomputing Center (PSC).

How XSEDE and PSC Helped:

Oakes began by testing cells' ability to adhere to substrates of different rigidities in the lab. He found two interesting things—one new, another unexpected. First, the cells didn't just test the substrate and move. The interaction of the lamellipodium with the substrate was "biphasic"—the cell used the lamellipodium to tug in two steps. If the substrate gave way too much, nothing else happened. But if the substrate tugged back, the cell pulled harder, the first step in moving.

The team's other finding upended what many scientists had expected about how the lamellipodium worked. Most had thought that the cell would pull on the substrate using myosin, a protein that serves as a tiny motor within the cell. But Oakes found that when he added a drug that prevents myosin from using ATP, the cell's basic fuel, the two-step tugging process still happened. A motor other than myosin had to be at work. Further lab testing pointed to the interaction between integrin, a protein that sits in the membrane that surrounds the cell, and the substrate. Integrin serves as the anchor of the cell to the substrate. But it does not run on ATP like myosin does. Instead, the energy integrins use to tug comes from binding and unbinding the substrate.

The scientists wanted to understand how the cell could use integrin to sense surface stiffness. Bidone created a virtual lamellipodium in the XSEDE-allocated Bridges system at PSC. She then changed, in tiny increments, how the different components of the cell behaved. She needed to simulate about 300 seconds of lamellipodium behavior in three dimensions, under thousands of different assumptions for how the components worked. Each simulation took three to four hours to compute, with about a two-fold speedup over other available computers. Bidone's computation used about 2,000 of Bridges' computational "cores"—by comparison, most top-line laptops have 4 cores. Bidone got off to a great start in part because of the online tutorials on, which helped her figure out how to set her calculations up successfully.

"A ‘catch bond' is a bond that strengthens as you pull on it. Depending on the applied force on the bond, it actually stiffens up and gets stronger, but above a certain threshold of force the bond disassembles. Tamara's simulations showed that, through this strengthening and weakening of the integrin binding, the substrate modulates its binding and unbinding, and very conclusively the simulations fit the experimental data. I think that's the key of where computation was valuable in this work. It showed the importance of the integrin catch-bond interaction with a substrate. Without computation, the interpretation of the experiments would be a lot more difficult."—Gregory Voth, University of Chicago

The cells' sensing of the surface rigidity, Bidone discovered, depended on a "catch-bond" mechanism. That means that the harder the surface tugged back, the longer the connections holding the integrin and fibronectin network together persisted. The virtual lamellipodia in her simulation recreated the real cells' behavior, duplicating the two-step process perfectly. It also reproduced the behavior of mutant cells with altered catch bonds. The team reported their results in the journal Proceeding of the National Academy of Sciences USA in March. Next, the scientists plan to study whether simple nudges by chemicals or other means might be used to direct whether cell movement happens or not. It's a first step toward drug therapies targeting cell movement in cancer and other disease processes.


Microscope images of a cell trying to get a foothold on a soft (A) and hard (B) substrate. The cell puts out a lamellipodium that grips the hard substrate, but only tests the soft one.

Previous years' ECSS seminars may accessed through these links:





Content with tag biology .

There are no results.