There is a pleasing symmetry to Sotirios Karathanasis’s career. As a young chemist from Greece, newly arrived in the United States, he met biology as a foreign country twice over. The science was new and so was the language. He worked through textbooks with a Greek-English dictionary beside him, translating his way toward a doctorate in biochemistry. Decades later, after Harvard Medical School and senior research posts at three pharmaceutical companies, he chose another translation problem: how to turn the obscure signals of natural molecules into structures that drug hunters can understand.
The first translation made a scientist. The second brought him to Enveda, a biotechnology company that combines metabolomics, automated experiments and machine learning to search living chemistry for possible medicines. Karathanasis joined as chief science officer and is now listed as a founding fellow. The title has changed; the preoccupation has not. His work keeps returning to the distance between information and use.
A gene sequence is information. A mass spectrum is information. A promising assay result is information. None is a medicine. Between the signal and the useful object stands a long sequence of judgment calls. Karathanasis has spent his working life in that sequence.
A dictionary, then a laboratory
He was born in Greece and studied chemistry at Aristotle University of Thessaloniki. After moving to America, he earned a Ph.D. in biochemistry at the University of Georgia. The reported image of those early years is wonderfully unfashionable: no instant translation, no clever app, only a student moving back and forth between a biology text and a dictionary until the words yielded their meaning.
Postdoctoral work took him to Harvard Medical School, where he later became an associate professor. His research centered on metabolism, especially the genetics and molecular biology of lipoproteins. In the 1980s, he co-authored work on the organization and mapping of human apolipoprotein genes, the instructions behind proteins that move fats through the bloodstream. The papers belong to an era when finding and sequencing a gene could itself be a substantial expedition.
This was not abstract mapmaking. The appeal lay in connecting molecular mechanisms to cardiovascular biology. The American Heart Association gave him an Established Investigator Award in 1985. His bibliography grew across molecular biology, lipid research, genetics and pharmacology. Research profiles now count more than 100 publications and thousands of citations. Numbers are blunt instruments, but these describe sustained attention rather than a brief encounter with a fashionable subject.
The useful discomfort of industry
Academia rewards the question that reveals something true. Drug development adds an impolite follow-up: can the truth survive contact with a molecule, a development plan and an organization? A cardiologist friend whom Karathanasis admired moved into industry. The example helped draw him in as well.
He became director of cardiovascular pharmacology at Pfizer Global Research & Development, then chief science officer for endocrine and cardiovascular research at Lilly Research Laboratories. At AstraZeneca, as vice president and head of biosciences, he led roughly 250 scientists in Gothenburg. His remit crossed locations, disciplines and the borders between biology and chemistry. The job was no longer only to understand a mechanism. It was to arrange people, evidence and resources so that a promising mechanism had a chance to become a program.
That requires a temperament comfortable with both microscopic detail and institutional scale. One colleague’s story supplies the comic footnote. At a pharmaceutical company, Karathanasis reportedly accumulated so much unused leave that he was ordered to take a vacation. He returned in track pants with a stack of papers and worked from someone else’s office. As disguises go, it lacked subtlety. As evidence of appetite, it was impeccable.
“The chemical diversity of nature is much smarter than what most companies use.”Sotirios Karathanasis
The anecdote is charming because it is excessive. It also hints at a constant in a career that moved through very different institutions. Karathanasis likes the work. Public biographical notes mention history and finance among his interests, pursuits united by their fondness for systems, incentives and consequences. Drug discovery offers all three, with the additional inconvenience that the system is alive.
Nature has a library problem
Natural products have supplied medicine with celebrated starting points, but the romance of the forest can obscure the nuisance of the flask. A leaf, fungus or tissue sample contains a complicated mixture. Researchers must separate its components, identify their structures, measure their activity and then obtain enough material to keep working. Conventional methods can be slow, expensive and dependent on scarce expertise.
Mass spectrometry helps by breaking molecules into charged fragments and recording their patterns. Yet a spectrum is closer to the scattered pieces of a sentence than to the sentence itself. Enveda’s models aim to infer chemical structure and biological relevance from those patterns at scale. The company calls the resulting platform PRISM. Its premise is that living systems contain an enormous collection of chemistry shaped by evolution, while scientific catalogues cover only a small part of it.
The translation chain
Karathanasis found the proposition persuasive because it joined an old source of medicines to tools that could ease its old bottlenecks. He pointed to metformin, statins and aspirin as reminders that nature-derived chemistry has already created durable families of drugs. The opportunity was not merely to admire natural diversity, but to make more of it searchable.
In 2023, while serving as Enveda’s chief science officer, he described the company’s work as a way to address difficult targets and basic disease processes. That year Enveda publicized MS2Mol, a transformer model designed to predict molecular structures from mass spectra. Karathanasis framed the ambition in clinical terms: identify promising chemistry quickly enough that it can enter the familiar, unforgiving machinery of drug development.
The veteran and the model
AI drug discovery invites a mischievous misunderstanding: if the model is clever enough, perhaps experience becomes optional. Karathanasis’s presence suggests the reverse. A model can enlarge the set of visible candidates. It cannot make every downstream choice. Which assay reflects the biology? Which signal deserves repetition? Which structure is chemically workable? Which liability will become expensive later? The wider the search becomes, the more valuable disciplined rejection may be.
His career supplies a long memory for such choices. He has watched molecular genetics mature, nuclear receptors become drug targets, regenerative medicine test the conventions of pharmaceutical development, and computation move from supporting role to center stage. In 2014 he argued that regenerative medicine would require pharmaceutical research to adapt its traditional habits. At Enveda, the adaptation concerns the front end of discovery: where useful molecules come from and how scientists recognize them.
Chemistry training becomes a biochemistry doctorate, learned across a language barrier.
Postdoctoral research grows into a faculty career focused on metabolism and lipoprotein genetics.
Scientific questions acquire portfolios, development constraints and teams numbering in the hundreds.
Natural chemistry, mass spectrometry and machine learning converge in a new search system.
The company around him has moved from premise to clinical-stage programs. In September 2026 Enveda announced a $311 million financing after reporting early clinical results from medicines discovered with PRISM. Patent publications in 2026 also named Karathanasis among the inventors on Enveda work involving isoandrographolide analogs. Neither event makes the path from nature to medicine short. They show that the path can be walked.
Karathanasis now holds the title of founding fellow, a role that suits the shape of the story. A fellow is permitted to range, to advise, to carry institutional memory without pretending that memory is the same thing as certainty. His useful gift may be knowing how many translations still remain after a machine offers an answer.
The sentence inside the spectrum
The dictionary from his student years makes an irresistible metaphor, but it should not be polished until it loses its grain. Learning a language word by word is frustrating. Drug discovery is frustrating in much the same way. Meaning arrives provisionally. Context changes it. False friends abound. Fluency is earned through error.
What changed across Karathanasis’s career is the quantity of text available to read. Early molecular biology opened genes to inspection. Industrial screening widened the field of compounds. Metabolomics now produces immense records of chemistry from living systems. Machine learning can propose meanings at a pace no dictionary-bound student could match.
But speed is not comprehension, and a structure is not a drug. The final translation still belongs to teams that can join computation to experiments, chemistry to biology, and curiosity to the patience of development. Karathanasis has worked on both sides of those joins. He began by learning the words. He stayed for the sentences.