THPharm already has a drug in late-stage trials. Now it’s using AI to work backwards and find out what else that drug might be good for.
Drug development usually runs in one direction: pick a disease, search for a molecule that treats it, spend years finding out if you’re right. THPharm, a South Korean biotech focused on metabolic disease, is trying it the other way around. Start with a drug that already works; let an AI system figure out everything else it might treat.
The company announced this month that it’s partnering with Insilico Medicine, one of the more recognizable names in AI drug discovery, to run that reverse approach on THP-001, THPharm’s lead drug candidate, which is currently in global Phase 3 trials [1]. The plan is to feed everything already known about how THP-001 behaves in the body into Insilico’s AI platform, PandaOmics, and see what diseases, biological pathways or early warning signs pop out that nobody was originally looking for.
Starting from the answer, not the question
Think of it like this: instead of asking “what drug treats heart failure,” THPharm is asking “we already know this drug does X, Y and Z in the body, so what else might benefit from that.” The AI combs through gene activity, known biological pathways and scattered scientific literature looking for patterns. It’s the kind of pattern-matching a human research team could theoretically do by hand, but realistically wouldn’t, because the relevant clues are often buried across dozens of unrelated studies that nobody thinks to connect.
So far, the two companies have pulled together 16 public datasets covering 269 samples to build the analysis, focused on how THP-001 changes gene expression after treatment. The candidates under consideration right now are heart failure, fatty liver disease and obesity – three of the biggest problems in metabolic health. Rather than chasing all three at once, THPharm says it’ll prioritize whichever shows the strongest, most repeatable signal in the AI analysis, then test that one in cell and animal models before it goes anywhere near a human trial.
Why skip straight to Phase 3 data
The expensive part of finding a new use for a drug is usually proving, from scratch, that it’s safe. THPharm’s bet is that it can skip a chunk of that risk by starting with a drug that already has real human safety data behind it, since THP-001 is already deep into a Phase 3 trial. That’s a leaner, cheaper way to build a pipeline, and it matters more right now than it might have a few years ago, with biotech funding tighter and investors wanting proof before they write big checks.
“Artificial intelligence enables us to extract far more value from existing biological data than was previously possible,” said Alex Zhavoronkov, PhD, Founder and CEO of Insilico Medicine. “We’re excited to support THPharm’s reverse engineering strategy with PandaOmics, helping connect pharmacological effects to disease biology and identify promising new therapeutic opportunities.”
“We will build a capital-efficient drug development model that selects clinically valuable pathways by linking confirmed pharmacological effects with disease mechanisms, ultimately creating multiple pipelines from a single asset,” said Tae Hee “Theo” Han, CEO of THPharm Corp.
Worth being clear-eyed here: this is still early, mostly computational work. None of the candidate diseases have been tested in animals yet, and an AI flagging a possible connection is a hypothesis, not a result. It still has to survive actual lab testing before it means anything.
For the longevity sector
Metabolic disease is more tangled up with aging than people tend to realize. Fatty liver disease, obesity, and heart failure don’t just shorten life on their own; they speed up almost everything else that comes with getting older, from cognitive decline to frailty to cardiovascular disease. A single drug that could meaningfully help with more than one of those conditions would be a big deal, medically and financially.
However, the real story is about a shift in how longevity-relevant medicines might get found in the first place. If tools like PandaOmics can reliably mine data that already exists for hidden therapeutic value, it could shrink the gap between “drug that treats one thing” and “drug that helps with several age-related conditions at once.” That’s the kind of efficiency longevity medicine is going to need if it wants to keep pace with a population that’s aging faster than the drug pipeline is currently built to handle.
[1] https://insilico.com/news/h9zeln8cc1-insilico-medicines-pandaomics-enables-ai