Why Souptik Barua prefers 'good enough'

ACTIVATE helped Rice alumnus define his work as a data scientist.

Souptik Barua

“Perfect is the enemy of good enough,” said Rice University alumnus Souptik Barua ’19 as he describes his method for publishing multiple research papers within a relatively short period of time.

Barua, who graduated from Rice with a Ph.D. in electrical and computer engineering, uses data science to derive novel clinical insights from digital health data that can improve disease understanding and provide new treatment and prevention avenues.

“My research on medical images and sensor data — which are stored digitally as numbers — uses techniques to convert those numbers into information that clinicians can use for earlier diagnosis of diseases like diabetes and cancer.”

Returning to the topic of publishing papers reasonably quickly, he said, “Is it more important to perfect my paper for publication in a renowned journal or to send off a ‘good enough’ paper to a less prestigious journal and get the information more quickly into the hands of the people who need it most?”

Like many of his peers, Barua started graduate school thinking academic success depended greatly on publishing one’s research in premier journals such as Nature or Science. He credits his mentors for changing his initial ideas about submitting his work only to the very top journals. They encouraged him to send his papers to a broader range of publications because they cared deeply about advancing the field they were working with Barua on.

“My mentors pointed out that the problem is the most important thing, not the reputation of the journal where it is published. If you feel what you are working on can be critical to the diagnosis or treatment of a disease, that is more important than your accolades,” he said.

“It was a big relief for me that my mentors never put any pressure to publish only in ‘Nature’ or ‘Science.’ Working without that stress allowed me to do my best work.”

Barua’s best work is achieved methodically. First, he plots a timeframe to familiarize himself with the problem and the data before attempting to find solutions.

He said, “I spend a reasonable amount of time in my research getting to know the problem well and studying the data I have. Then, after several months of just working with the data, that gives me a few ideas at once, and perhaps two or three can become papers.

“I’d advise researchers wanting to try this method not to write while you are spending time understanding the problems, solutions, and ideas. Once you’ve figured those out, then you can write back-to-back about those ideas and send out several papers.”

Barua said he is an interdisciplinary researcher, so scientists or engineers in overlapping fields can build on his ideas. When they extend his initial work, he benefits from co-authorship of a new paper. But he is also interested in collaborative work with clinicians, which helps build their trust in his findings. In addition, their feedback can guide his subsequent steps.

“I hope my work can discover new digital signatures on biomarkers that can tell us about the disease before it progresses. Particularly for chronic diseases, earlier detection and treatment extends the patient’s outcome. My research helps with the prediction of onset or identifying early warning signs of the disease.

“We may also be able to generate novel insights about a disease, perhaps go beyond what we already know about digital signatures and discover how they progress; that is shedding more light on the biological mechanisms of the disease.”

He believes collaboration between data scientists and clinical experts from the start will be the key to fully utilizing data in ways that best benefit patients. By talking to his clinical collaborators and asking which parts of the data are more important than others or where they draw their best insight, Barua can design more appropriate data science frameworks. Then, his next set of approaches can better pinpoint those areas, making the results easier to interpret clinically.

Barua said, “The easier we make it to incorporate the findings from data science or machine learning into clinical practice, the more likely we are to build trust in that approach. This would amplify the belief clinicians have in our results and will incorporate it in their treatment plans.”

His passion for data-driven understanding of disease to improve diagnosis and treatment was endorsed with a Future Faculty Fellowship in the George R. Brown School of Engineering at Rice. The FFF program was created to support Ph.D. candidates and postdoctoral researchers on their path to academic careers as tenure-track faculty.

“I knew what I wanted to do,” said Barua. “I’m a data science researcher trying to understand and improve health applications.”

The best place to achieve the most significant impact for his work appeared to be a faculty position where he could both teach data science applications to future scientists and engineers and continue advancing his research into early disease markers.

“Several FFF workshops were targeting different parts of the faculty search, including the development of a powerful letter of application. We also learned to write support materials and improve components like our teaching statement, research statement, and diversity statement,” he said.

The FFF coaches helped the participants better understand the benefits of broadening their search and taught them how to read calls for applications and unpack the content to determine specific areas in which the search committee would be interested. Barua said the “unpacking” lesson prompted FFF participants to consider opportunities they might have otherwise overlooked.

“Learning from experts in the ACTIVATE team, we identified key areas where the hiring committee would probe, and we made sure we covered the high notes in our applications: define who you are, what you want to do, and what vehicles you will use to get there.”

Barua then took advantage of workshops tailored to faculty interviews, including tips for talking with faculty members and how to speak to a dean —who would be more interested in how candidates tied into the university’s initiatives. The FFF participants learned ways to improve their job talk itself (how candidates present their research), decide what to include and what to leave out, answer questions, and what questions to ask the hiring committee in return.

He said, “The FFF workshops were very illuminating. It was intriguing to look at the process from the perspective of candidates interviewing potential employers. As postdocs, we tend to get very deep in our research. The opportunity to look at other areas that influence finding a faculty position was genuinely beneficial.

“Our FFF experiences were also dynamic and interactive. We didn’t just sit in a lecture listening to what we should do. Instead, they had us role play: ‘Now I’m the dean, how would you describe your work and goals to me succinctly?’ All of the sessions helped me prepare for each aspect of the application process with confidence.”

When asked for his final thoughts, Barua said while his work involves analyzing enormous amounts of data, he recognizes the importance of where the data came from.

“The most important part of my research is keeping in mind that the data isn’t just numbers. This is a person with an illness or the potential for developing an illness,” he says, “And therefore I constantly ask myself: how will my data-driven technology benefit that person? Will my technology be inclusive of all people?”

He hopes that if we keep the big picture in mind while designing data science or AI solutions, we will progress faster in our quest for improved clinical outcomes for all.

This article was originally published for the ACTIVATE Engineering Communication program.