The creator of star drug-discovery algorithms for Pharma brings big data—and big ideas—to campus
BY SEAN MARKEY
PHOTOGRAPHY BY MATT FURMAN
NORWICH RECORD | Summer 2020
If there has ever been a moment when the world needs data science and data scientists, it is now. “The world has been shutting down. People are unable to travel. Cities and countries are entirely [closed] down,” says Assistant Prof. Ahmed Abdeen Hamed, PhD. “The only thing that’s going to save us right now is data science.”
By save, the data analytics wunderkind means save us from the coronavirus pandemic. Researchers sequenced the genome of SARS-CoV-2 just 43 days after it first emerged in humans. But only by using data science. And data science will empower every breakthrough and hoped-for breakthrough as scientists race to find antiviral treatments, cures, and, God willing, a vaccine—thus safeguarding the lives and livelihoods of nearly every human on Earth, all 7.8 billion of us, from personal extinction.
Hamed is a computer scientist who joined the Norwich faculty last summer to stand up new programs in data science. Before he arrived, he spent a decade and a half working on the forward edge of the revolution in big data, machine learning, networks, and data analytics. Among his talents is the ability to develop complex network algorithms that creatively mine huge data sets, from libraries of scientific research to billions of tweets in the Twittersphere, to make novel discoveries.
He has an uncanny knack for hitting on the next big idea, followed by a half-dozen more. Ones that leave a trail of highly regarded academic papers if not industry patents in their wake. Then there are the accolades that defy easy categorizing. Like the time in 2016 that the editors of Fast Company magazine named him to their annual list of the 100 “Most Creative People in Business.”
What makes Hamed’s field both possible and necessary is the ocean of data, vast and growing by the second, that encircles our world. Every device we own, from our computers to our smart phones to our cars, trails clouds of data. Every day we create billions of social media posts, to say nothing of scientific research, which is a data-generating engine running at warp speed. A case in point: In just four months, scientists published nearly 15,000 studies related to COVID-19. Which is to say, there is more information than ever. If only we could find the right bit, the right insight to make new discoveries.
“You could be working on something very specific to your field, to your grant, to your research,” Hamed says, “and you could easily overlook so many different things that overlap if you don’t have some sort of a machine intelligence of what people are doing around you.” Enter the world of big data and data analytics.
Over the course of his career, Hamed has hopscotched back and forth between industry and academia. The diversity of his work experience, coupled with his intellectual restlessness, has become, in hindsight, a strength. “I was always creating something, leaving it behind, and moving on to the next level,” he says. “I didn’t want to stop.”
Partway into his PhD program at UVM, Hamed hit the pause button to work in private industry for a tech start-up that develops software and websites for in-store and online grocery sales. While he was there, Hamed wrote an algorithm that “played the guessing game” of what you planned to make for dinner, based on the items in your cart. The payoff came when his algorithm could correctly guess “pot roast” and spot that you forgot to buy the rosemary.
Soon after, Hamed resumed his PhD, where he became the protégé of Xindong Wu, PhD, a pioneer in data mining. One of Hamed’s early PhD projects was the creation of an AI-powered social bot that could chat with smokers on social media. A rare early example of a benign social bot, it was designed to recruit smokers to cessation programs.
After that project, Hamed says he grew increasingly interested in Twitter and how its millions and billions of posts could be mined for insights. At the time, medical marijuana was in the news. Hamed was curious about how it might adversely interact with other medications. Mostly he was just curious. He says he didn’t have any preconceived ideas about what he was looking for or even hoping to discover. “I had a landscape that I wanted to walk around and see what the landscape was going to tell me,” he says.
Working with colleagues at UVM and the Albany College of Pharmacy, Hamed created software to systematically search billions of tweets in the massive data repository of the Twittersphere. His “aha” moment came when his algorithm revealed links between “medical marijuana” and the hashtag “#Alzheimer’s.” Using the tool to dig more deeply, Hamed found tweets which suggested that taking Ibuprofen stopped the memory loss associated with medical marijuana. Such loss is an early risk factor for Alzheimer’s disease. Hamed and his colleagues went on to develop the tool as a means to search social media posts to discover unknown drug interactions. It was that work that prompted the editors of Fast Company to name Hamed to their list of “Most Creative People in Business.” In his typical stepping-stone fashion, it also laid the foundation for future breakthroughs.
“His doctoral research is a great example of exactly why big data can create knowledge,” says Aron Temkin, dean of NU’s College of Professional Schools, which houses the School of Cybersecurity, Data Science, and Computing where Hamed teaches. “If you gather anecdotal evidence from a dozen people, it’s just storytelling. But if you gather it from millions of people, right, then it’s not surprising that you can start to see patterns.”
“I don’t want people to think that I’m bragging,” Hamed says. “I mean, it’s cool. [But] I don’t want to gloat. I want science. But the point is, I’m coming back with all of this richness and all these ideas. And now I’m at Norwich, and we’re creating a new data science program.”
Last fall, Hamed taught the program’s first data analytics course, adding a second course in Python, the go-to coding language for data analytics, in the spring. He is already designing and planning additional courses in machine learning, algorithms, and artificial intelligence.
Hamed radiates and induces energy and enthusiasm for the field, says faculty colleague Matt Bovee, PhD. The director of the School of Cybersecurity, Data Science, and Computing audited Hamed’s first data analytics class last fall. “Every time I’ve seen someone come into [his] class and start asking questions about their own particular area, they get really excited about what they can do,” Bovee says.
Sarah Eriksson ’19 is one such student. A cybersecurity major from Danbury, Conn., who graduated in December, she was a high school freshman when the Sandy Hook Elementary School massacre happened in Newtown, Conn., 20 minutes from her home. That experience inspired her to work on a data analytics project in Hamed’s class to trawl Twitter to identify potential school shooters. Eriksson, who now works in cyber-security for Honeywell, says she loved the course and Hamed’s committed teaching style, pointing to his Saturday “data-thons” and instant replies to email—whatever the time, even at midnight.
Dean Temkin says, “We were really looking for somebody who didn't just want to teach a bunch of specialists, but could be enthusiastic and excited about all of the ways that big data and data analytics could be used as a tool. Ahmed’s fantastic for that, because his enthusiasm for the subject’s really palpable. He thinks very broadly. He’s very interested in writ¬ing courses that can be reached by students in many different disciplines.”
Some of the richness that Hamed brings to campus is tied to the three and half years he spent in a Boston-area research office of Merck, one of the world’s top pharmaceutical companies. His official title was senior applied computer scientist. But really his job was to invent. Shortly after he landed, Hamed learned about the company’s “Better Molecules Faster” initiative. Most of us can be forgiven if we have only the vaguest idea of what drugs are. But for pharmaceutical researchers, the work is really about discovering new molecules. Each one is a potential new treatment for a disease or illness. Other molecules might be repurposed in new ways. An example from today’s headlines would be Remdesivir, an experimental drug that failed to treat Ebola but is now being used as a moderately effective antiviral against COVID-19.
Intrigued by the task of discovering molecules, Hamed saw a potential way to leverage his skills: Suppose he wrote an algorithm that enabled scientists to search private company and public academic research databases for every known drug molecule and rank them based on their potential to treat a particular disease or condition? Hamed knew that, as a “mere computer scientist,” working with the medical literature was beyond his capabilities. So he teamed up with Agata Leszczynska, a pharmaceutical scientist and PhD based in Prague. For the next two years, the pair collaborated from different time zones, video conferencing weekly as they worked on their novel molecule-ranking engine, which they dubbed MolecRank.
How to sort the most promising molecules proved tricky. But Hamed and team eventually landed on the idea of specificity. The more specifically a molecule affected a specific organ or cell type or cell line or even a gene, the greater its potential as a possible treatment. The converse was also true.
“The idea was to identify the molecules based on how specific they are,” Hamed said. “If it targets the wrong cell, then you have an undesirable side effect. It could kill the person.”
To determine whether or not a potential drug is dangerous, pharmaceutical companies can spend millions of dollars designing tests and clinical trials that can last three to five years. “We thought we could do a preliminary analysis just from [existing] publications and what people have [already] done.”
It took two and a half years to prove, but they were right. MolecRank opened new avenues of knowledge discovery to researchers. Hamed was credited as first inventor on its patent. The pandemic spotlights the algorithm’s timeliness and import, he says today, especially for drug repurposing and the quest to find existing drugs that can effectively treat the coronavirus.
While he was at Merck, Hamed was tasked with another Mensa-sounding project—serving as the quantum computing lead for problems known to be intractable. The field involves a set of as-yet unanswered, fundamental questions about the chemical structures and substructures of molecules. The problems are so challenging they “cannot be solved on the most powerful computers that exist on Earth,” Hamed says. For now, those questions remain unanswered.
Hamed has always wanted to make the world a better place and his affinity has always been for application over theory. Shortly after he finished his master’s degree at Indiana University, a friend lent him a DVD set of Planet Earth, the BBC nature documentary series hosted by David Attenborough, one the world’s leading naturalists. Hamed fell in love with the world’s biodiversity and grew angry that it was in such peril. So he did what he often does. He had an idea. Suppose he wrote an algorithm? One that leveraged Twitter to generate a real-time status report for endangered species? His plan changed. But its spark speaks to his ethos. “I do not care about symbols, and I do not care about designing the next programming language that makes people really excited about being programmers. I am interested in actual problems.”
More than a decade later, Hamed still spends much of his time thinking about problems. But not all of it. He finds time to play the oud, the Middle Eastern stringed instrument he first taught himself to play 25 years ago and sometimes plays for his teenage daughter. He is also a rabid fan of English Premier League soccer, specifically, Liverpool, F.C., the club that paid £36.9 million (about US$45 million) in 2017 for Mo Saleh—the phenom Egyptian forward who, according to Wikipedia, is among the greatest players in the world and who, according to Hamed, is more simply “a legend.” Hamed himself still plays pickup soccer nearly every week. Or used to, before the pandemic. He looks forward to playing again. When it’s safe again, when data science has saved us all.