Skip to content

Commit

Permalink
Merge Split Branch
Browse files Browse the repository at this point in the history
  • Loading branch information
Automatic preview publish committed Nov 14, 2023
2 parents 5fa58ec + 34f9c4f commit 9890c4e
Show file tree
Hide file tree
Showing 3 changed files with 1,431 additions and 1,411 deletions.
15 changes: 7 additions & 8 deletions preview-update-submodules/featured-users/2023-11-13-spalding-lab.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -783,7 +783,7 @@ <h1>Plant physiology researchers used high throughput computing to remedy resear
</div>


<div class="mt-4 fs-5"><p><strong>Once they began using HTC resources, the Spalding Lab was able to increase their efficiency in data collection and analysis and develop new methodologies.</strong></p>
<div class="mt-4 fs-5"><p><strong>HTC resources increased the efficiency of the Spalding group’s data analyses, which enabled an increase in the scope of their research.</strong></p>

<p>Enhancing his research with high throughput computing was a pivotal moment for University of Wisconsin–Madison molecular plant physiologist Edgar Spalding when his
<a href="https://spalding.botany.wisc.edu/">research group</a> started using it in 2006. Over the past five years, the research group has used approximately 200,000 computing hours, including as a way to “develop the measurement algorithm and to process the tens-of-thousands of images automatically” of maize seedling root growth, Spalding says.</p>
Expand All @@ -795,18 +795,17 @@ <h1>Plant physiology researchers used high throughput computing to remedy resear

<p>Spalding’s research group was studying Arabidopsis plant populations with genetically diverse members and tracking their response to light or gravity due to a mutation — one seedling at a time. Since Arabidopsis seedlings are only a few millimeters tall, Spalding says his research group found that obtaining high-resolution digital images was the best approach to measure the direction of their growth. A computer collected images every few minutes as the seedlings grew. “If we could characterize this whole genetically diverse population, we could use the powerful techniques of statistical genetics to track down the genes affecting the process. That meant we now had thousands and thousands of images to measure,” Spalding explains.</p>

<p>The thousands of digital images to measure created a bottleneck in Spalding’s research. That was before he led an effort with <a href="https://chtc.cs.wisc.edu/">Center for High Throughput Computing (CHTC)</a> Director Miron Livny, other plant biologists, and computer scientists to develop a proposal for a competitive National Science Foundation grant that would produce cyberinfrastructure to support plant biology research. Though the application wasn’t successful, the connections Spalding made from that meeting were meaningful nonetheless.</p>
<p>The thousands of digital images to measure created a bottleneck in Spalding’s research. That was before he led an effort with the <a href="https://chtc.cs.wisc.edu/">Center for High Throughput Computing (CHTC)</a> Director Miron Livny, other plant biologists, and computer scientists to develop a proposal for a competitive National Science Foundation (NSF) grant that would produce cyberinfrastructure to support plant biology research. Though the application wasn’t successful, the connections Spalding made from that meeting were meaningful nonetheless.</p>

<p>Speaking with Livny at the meeting — from whom he learned about high throughput computing — helped Spalding realize the inefficiencies of his group in analyzing thousands of seedlings. “[O]ur research up until that point had been focused on one seedling at a time. Faced with large numbers of seedlings to do a broader scale of investigation meant that we had to find computing methodologies that matched our new data type, which was tens of thousands of images instead of a couple of dozen. That drove our need for a different way of computing,” Spalding describes.</p>
<p>Speaking with Livny at the meeting — from whom he learned about the capabilities of the HTC approach that was pioneered on our campus — helped Spalding realize the inefficiencies of his group in analyzing thousands of seedlings. “[O]ur research up until that point had been focused on one seedling at a time. Faced with large numbers of seedlings to do a broader scale of investigation meant that we had to find computing methodologies that matched our new data type, which was tens of thousands of images instead of a couple of dozen. That drove our need for a different way of computing,” Spalding describes.</p>

<p>When asked about which accomplishment using high throughput computing (HTC) was most impactful, Spalding said “the way we measure yield-related features from maize ears
and several thousand kernels has had a large impact.” The Spalding Lab had others from around the world asking for their help in making these measurements. “In many cases, we can use our workflow [algorithms] running on CHTC to process their images of maize ears and kernels and return data that helps them answer their scientific or crop breeding questions,” Spalding says.</p>
<p>When asked about which accomplishment using HTC was most impactful, Spalding said “The way we measure yield-related features from maize ears and several thousand kernels has had a large impact.” Others from around the world began asking for their help with making similar measurements. “In many cases, we can use our workflow [algorithms] running on CHTC to process their images of maize ears and kernels and return data that helps them answer their scientific or crop breeding questions,” Spalding says.</p>

<p>The Spalding Lab did not need to adjust the type of data they collected. Rather, using HTC changed the way they created tools to analyze the data. Today, Spalding says his research group continues to use HTC in three ways: “from tool development to extracting the features from the images with the tool that you developed to applying it in the challenge of statistically matching it to elements of the results to elements of the genome.” As his team became more experienced in writing new algorithms to make measurements, they realized that HTC was useful in developing new methodologies; it was more than just increased computing capacity.</p>
<p>Since the goals of the experiments determine the type of data the researchers collect, they did not need to adjust the type of data they collected. Rather, adopting the HTC approach changed the way they created tools to analyze the data. Today, Spalding says his research group continues to use HTC in three ways: “from tool development to extracting the features from the images with the tool that you developed to applying it in the challenge of statistically matching it to elements of the results to elements of the genome.” As his team became more experienced in writing new algorithms to make measurements, they realized that HTC was useful in developing new methodologies; it was more than just more automation and increased computing capacity.</p>

<p>In other words, HTC is useful as both a development resource and a production resource. Making measurements on seedlings and then matching processes to the genome elements that control those processes involved a lot of computation. “We realized that statistical modeling of the measurements from the biology to the genetic information in the population also benefited from high throughput computing.” HTC in all these cases, Spalding elaborates, “was beneficial and changed the way we work. It changed the nature of the questions we asked.” In addition to these uses of HTC, the research group’s uses of machine learning (ML) also continue to become a bigger part of the tool development stage and in driving the methods to train a model to recognize a feature in a seedling.</p>
<p>In other words, HTC is useful as both a development resource and a production resource. Making measurements on seedlings and then matching processes to the genome elements that control those processes involved an ever-growing amount of computing capacity. “We realized that statistical modeling of the measurements from the biology to the genetic information in the population also benefited from high throughput computing.” HTC in all these cases, Spalding elaborates, “was beneficial and changed the way we work. It changed the nature of the questions we asked.” In addition to these uses of HTC, the research group’s uses of machine learning (ML) also continue to become a bigger part of the tool development stage and in driving the methods to train a model to recognize a feature in a seedling.</p>

<p>Spalding has also spoken at the <a href="https://osg-htc.org/user-school-2023/">OSG School</a> in the past. Spalding emphasizes that students “should not do something because they think computing will be a bottleneck. There are ways to bring the computing they need to their problem and they should not shy away from a question just because they think it might be difficult to compute. There are people like the CHTC staff that can remove that bottleneck if the person’s willing to learn about it.”</p>
<p>Spalding has also shared his HTC experience with the attendees of the annual <a href="https://osg-htc.org/user-school-2023/">OSG School</a>. Spalding emphasizes that students “should not hold back on doing something because they think computing will be a bottleneck. There are ways to bring the computing they need to their problem and they should not shy away from a question just because they think it might be difficult to compute. There are people like the CHTC staff that can remove that bottleneck if the person’s willing to learn about it.”</p>

</div>
</div>
Expand Down
51 changes: 36 additions & 15 deletions preview-update-submodules/index.html
View file
Original file line number Diff line number Diff line change
Expand Up @@ -838,6 +838,8 @@ <h3 class="text-decoration-none m-0">
<div class="ms-n2">







Expand All @@ -853,19 +855,26 @@ <h3 class="text-decoration-none m-0">










<div class="">


<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-11-13-spalding-lab.html">
<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-11-10-chtc-workshop.html">


<div class="card border-0 ">
<div class="p-2 d-flex flex-column">
<h6 class="text-decoration-none mb-1">
Plant physiology researchers used high throughput computing to remedy research “bottleneck”
CHTC Launches First Introductory Workshop on HTC and HPC
</h6>
<div>
<span class="text-muted text-decoration-none">November 13, 2023</span>
<span class="text-muted text-decoration-none">November 10, 2023</span>
</div>
</div>
</div>
Expand All @@ -875,19 +884,22 @@ <h6 class="text-decoration-none mb-1">

</div>




<div class="">


<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-11-10-chtc-workshop.html">
<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-11-1-htcondor-european-workshop.html">


<div class="card border-0 ">
<div class="p-2 d-flex flex-column">
<h6 class="text-decoration-none mb-1">
CHTC Launches First Introductory Workshop on HTC and HPC
HTCondor European Workshop returns for ninth year in Orsay, France
</h6>
<div>
<span class="text-muted text-decoration-none">November 10, 2023</span>
<span class="text-muted text-decoration-none">November 1, 2023</span>
</div>
</div>
</div>
Expand All @@ -897,19 +909,22 @@ <h6 class="text-decoration-none mb-1">

</div>




<div class="">


<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-11-1-htcondor-european-workshop.html">
<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-10-30-david-swanson-awardees-2023.html">


<div class="card border-0 ">
<div class="p-2 d-flex flex-column">
<h6 class="text-decoration-none mb-1">
HTCondor European Workshop returns for ninth year in Orsay, France
OSG David Swanson Awardees Honored at HTC23
</h6>
<div>
<span class="text-muted text-decoration-none">November 1, 2023</span>
<span class="text-muted text-decoration-none">October 30, 2023</span>
</div>
</div>
</div>
Expand All @@ -919,28 +934,34 @@ <h6 class="text-decoration-none mb-1">

</div>

<div class="">


<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/featured-users/2023-10-30-david-swanson-awardees-2023.html">



<div class="">


<a class="text-decoration-none d-block hover-border rounded h-100" href="/web-preview/preview-update-submodules/news/htc23-concludes/">


<div class="card border-0 ">
<div class="p-2 d-flex flex-column">
<h6 class="text-decoration-none mb-1">
OSG David Swanson Awardees Honored at HTC23
Throughput Computing 2023 Concludes
</h6>
<div>
<span class="text-muted text-decoration-none">October 30, 2023</span>
<span class="text-muted text-decoration-none">July 14, 2023</span>
</div>
</div>
</div>


</a>

</div>
</div>



</div>

</div>
Expand Down
Loading

0 comments on commit 9890c4e

Please sign in to comment.