Research Is Messy
By Ken Chiacchia, PSC
XSEDE's Campus Champion Fellows can enrich Extended Collaborative Support Services projects, benefitting the PI's research, the fellow and the ECSS support expert, Pittsburgh Supercomputing Center bioinformatician Alexander Ropelewski told the December 2018 ECSS Symposium. But the best results require acknowledgement of the diversity of needs for each participant, clearly defined goals and managing expectations among subject experts who don't necessarily understand each other's fields.
"It has been very rewarding working with the Campus Champions fellows, and I would highly encourage anyone working in ECSS … to take advantage of the opportunity," said Ropelewski, who is Director of Biomedical Applications at PSC and an ECSS bioinformatics expert. However, "It's important to understand who's working on the project and what their viewpoint on bioinformatics is."
The Campus Champions Fellows is a competitive program for XSEDE Campus Champions, with four to five fellows selected annually for a one-year stint working with an XSEDE user's team. Its goal is to increase expertise on campuses by including Champions on ECSS projects. The fellow selects an ECSS project of interest, and then writes a proposal for approval by the PI and the ECSS staff.
Ropelewski described two successful CC fellows/ECSS collaborations, one at the University of Pittsburgh and one at Tennessee State University, an historically minority-serving institution. The two projects pose interesting similarities and differences, he said.
Brian Couger became a Campus Champion in 2014, while a graduate student at Oklahoma State University. Today he's an XSEDE Domain Champion for bioinformatics and Bioinformatics Specialist at OSU's High Performance Computing Center. Working with the University of Pittsburgh group led by PI Uma Chandran and Ropelewski, he undertook to mine human cancer sequence data in novel ways, to identify disease-relevant sequences in populations and sub-populations.
As part of the project Couger developed workflows for assembling, annotating and quantifying cancer sequence data. Thanks in part to his work the PI is a continuing XSEDE user, and has continued to collaborate with Ropelewski in other projects.
Chet Langin joined Campus Champions in 2017. A Research Coordinator at Southern Illinois University, he is "essentially the only person on campus providing computer support for research projects," Ropelewski said. The goal of his project, working with Suping Zhou at Tennessee State University, was to use de-novo assembly and annotation to identify and understand fungal and bacterial populations in the microenvironment of the goat rumen. The project is important because ruminant animals present a rich environment in which to discover previously unknown cellulose-degrading microorganisms.
Like Couger, Langin developed workflows for assembling and annotating microbial metagenomic data, using XSEDE-allocated resource Bridges at PSC. His work, however, provided a surprise, revealing that the datasets originally thought to be "fungal" were in fact overwhelmingly bacterial in origin. The work identified the five most prevalent bacteria in the rumen, which appear to be major actors in the goat's ability to digest grasses and other plants not digestible to humans—a major benefit of such livestock. The work resulted in a poster at the PEARC18 conference and what Hui Li, the Zhou lab member who worked on the project, described as "a great learning experience for all the students in the lab."
Ropelewski stressed the importance of setting expectations and meeting the goals of all the participants in a Campus Campions/ECSS collaboration. Participants will vary from biologists with modest computational experience to "bioinformatics engineers" whose interests are chiefly at the computational end—as well as the true "bioinformatics scientist" whose interest bridges the two worlds.
Ropelewski's suggestions for a successful collaboration include:
- At least one face-to-face meeting between the ECSS expert and the fellow; the PEARC conference is a perfect venue.
- Defining a successful outcome up front, including the fellow's expectations as well as the PI's.
- Integrating the fellow into regular meetings with the PI's team, again with as many face-to-face meetings as possible.
- Managing expectations throughout.
"I would say that each group struggles in some way," to understand the other's field of expertise, Ropelewski said. Computational experts can focus on beautiful, effective code without understanding how fundamental it is to lab scientists to quickly report their findings in the literature. Biologists may not understand that a year of work by a CC fellow is unlikely, by itself, to produce such a paper, let alone grants or patents. "Every bioinformatician is not equivalent … I think it's also important to manage expectations in the sense that research is messy."