On one side are scientists accusing a top Massachusetts Institute of Technology professor of claiming other researchers’ work as his own. On the other are colleagues rushing to his defense after a blistering and painstaking takedown. And it is all getting rather heated — and public.
At stake is not only the reputation of a scientist at one of the country’s most elite institutions, but also the possibility of a black eye for others in a burgeoning field of research.
The controversy is centered on work by Ram Sasisekharan, a professor of bioengineering whose algorithm-based methods were used to discover new medicines. Earlier this week, the private biotech firm Adimab published a paper asserting the treatments are almost identical to compounds previously described by other labs.
Contacted by STAT, three of Sasisekharan’s scientific peers said that the claims by Adimab were convincing and that it’s either impossible or highly unlikely that the antibodies are in fact the products of an algorithmic model. Two of them said his published papers were “misleading.” One said that the accusations against Sasisekharan, if substantiated, would mean his algorithms were “of minimal consequence.”
Two experts questioned the evidence against Sasisekharan and criticized Adimab for directly publishing its critique in an academic journal instead of raising concerns with him or the journals that published his work.
“I look at the paper and I say, ‘Where’s the beef?” said Dr. Peter Dedon, an MIT professor of biological engineering who has taught alongside Sasisekheran. “What’s going on here?”
In a new statement, Sasisekharan said the Adimab paper was “founded on a baseless conjecture and filled with entirely false claims.” MIT declined to comment.
Adimab, which sells antibody discovery services to drug companies, looked at two papers from Sasiskheran and his colleagues that claimed to use a computational approach to develop therapies. The antibody treatments the company examined — for influenza and Zika virus — each bore a striking resemblance to previously published compounds, according to Adimab. The telltale sign was the amino acid sequences, which form the backbone of an antibody, the company said.
Several of the outside experts who reviewed Adimab’s accusations and the original papers expressed serious doubts that Sasisekheran’s computer model could generate nearly identical sequences to those described by scientists using traditional methods.
“It’s statistically and physically impossible,” said Sachdev Sidhu, an antibody engineering researcher at the University of Toronto who was not involved in the analysis but who previously served on Adimab’s scientific advisory board. “There’s just no way that a sequence that was previously available to those authors would be mimicked by another algorithm. … Impossible, absolutely impossible. I’ll just put it that way.”
Adimab alleges that Sasisekharan’s influenza therapy — detailed in a 2015 Proceedings of the National Academy of Sciences paper — is very close to one published four years earlier in Science. The Zika antibody — which Sasisekharan’s lab described in Cell last year — is nearly identical to one described in Nature in 2016, according to Adimab.
Spokespeople for Cell and PNAS and said they were looking into the matter but declined to comment further.
In the instances cited by Adimab, Sasisekharan and his colleagues didn’t publish the amino acid sequences of the antibodies they claimed to have designed. The Adimab researchers found the sequences by looking for the resulting patent applications and checking them against GenBank, an open-access database of sequences.
Sai Reddy, an assistant professor at ETH Zurich who studies antibody engineering, said Adimab’s findings suggest that Sasisekharan’s work was “misleading.”
“It didn’t really accomplish what they said it did,” Reddy said.
The potential impact of the accusations could stretch beyond scientific publishing or academic research. Visterra — one of the companies Sasisekharan founded — was developing the flu antibody when it was acquired by the Japanese drug maker Otsuka for $430 million last year. And Sasisekharan received funding from the National Institutes of Health for the research behind both papers that have now been called into question. The grants were among dozens awarded to Sasisekharan by the NIH over the past two decades.
Dedon, the MIT bioengineer, said there are “big concerns” about Adimab’s paper. The company’s business depends on getting paid to discover antibodies of its own, giving it a commercial interest in discrediting Sasisekharan, he said.
“This is like corporate warfare being waged in an academic journal,” Dedon said.
He criticized Adimab for publishing its critique without first getting a response from Sasisekharan or the journals that published his work on influenza and Zika. The dispute should have been settled behind closed doors without publicly accusing Sasisekharan of deceit, he said.
Paul Schimmel, a professor of molecular biology at Scripps Research Institute in Florida, agreed. And he said Sasisekharan “has a plausible case” for having discovered the two antibodies independently.
“I really do think it’s a tempest in a teapot,” he said. “And it’s really damaging. It should never have been done this way.”
Tillman Gerngross, CEO of Adimab and co-author of the paper, said the idea that his company was motivated by commercial gain is “laughable.” Adimab doesn’t develop therapies of its own, and it has no partnerships in either influenza or Zika, he said.
Gerngross, who is also a professor of bioengineering at Dartmouth College, said Adimab chose to go public with its findings in order to allow the scientific community to consider the merits of its case. If anyone disproves Adimab’s work, the company will gladly take down the paper, he said.
Daisuke Kuroda, an antibody engineering researcher at the University of Tokyo, said there’s an easy way for Sasisekharan to prove the validity of the results: run the computational modeling again.
“If they claim that the two antibodies were purely computationally created, their results should be reproducible in another round of simulations with the same condition,” he said.
But if the results aren’t reproducible, the situation could prove to be a black eye for the field of antibody design, scientists said.
“It’s a disappointing development, and in my opinion, not representative of the field, which has made progress in recent years,” said Jeff Gray, a Johns Hopkins professor who studies antibody engineering.
Sidhu, the University of Toronto researcher, said he’s concerned that the alleged deception will give funding agencies and peer reviewers pause when they see research or proposals using computational methods.
“It’s literally stabbing other people in the back and making it harder for them to get funding,” Sidhu said.