Did you see the movie where Spider-Man files his taxes? Or the one where Wonder Woman sits on hold with her insurance company while her pasta water boils over? Or where Captain America finds blight on his tomato plants and drives to the county extension office where he spends fifty minutes with a seventy-four-year-old master gardener named Marlys then leaves with a handwritten note covering his soil composition, his watering schedule, and what Marlys calls “the mulching situation”?
No. Because the ordinary day-to-day doesn’t stand a chance next to the saving of the world.
We spend most of our lives in the ordinary. Not because we’re failing to reach the extraordinary, but because the ordinary is what holds everything together while we get there. It’s not the backdrop but the foundation. It’s what the story depends on, whether or not it gets any credit.
Drug Discovery Has a Storytelling Problem
Drug discovery runs almost entirely on ordinary days, punctuated by the moments that make the news: a new target gets identified, a compound shows promise, a trial produces results. Those moments get the headlines, press releases and keynote slots. What doesn’t get the same attention is the years of work behind those moments: the assays, the failed experiments, the redesigns, the slow accumulation of evidence that either holds up or doesn’t. That work has always been the majority of drug discovery.
Some of the most important work in drug discovery ends in a result nobody publishes, but a dead end isn’t a failure of the program. It’s the program working. The researcher who rules something out has learned something true. That knowledge travels forward even when it doesn’t make the headline because it can redirect the next hypothesis, narrow the next experiment or just quietly move things along. That work moves research forward without anyone announcing it.
The Shiniest Thing in the Room
Artificial intelligence is drug discovery’s latest extraordinary announcement, and the fanfare is legitimate. Most of the druggable proteome has never been touched. Of approximately 4,500 human proteins considered druggable, all approved drugs to date work through only 716 distinct targets. Drug hunters knew there was more biology to address but lacked a way to find and prioritize candidates at scale. AI is changing that. By scanning genetic evidence, biological networks and scientific literature at a scale no human team can match, AI is surfacing targets that were previously out of reach and ranking them by the strength of the evidence behind them.
Some of those nominations have already reached clinical trials. A novel target for idiopathic pulmonary fibrosis, identified by an AI platform, moved from nomination to Phase II in roughly eighteen months. A PIKfyve inhibitor for ALS, nominated with AI support, reached clinical trials but was discontinued in 2025 after it didn’t demonstrate sufficient efficacy. Both count. That’s how science moves. The candidate list is getting longer, and the targets on it are getting more specific. For diseases that have run out of obvious targets to try, that matters.
What’s not new is what happens next. Every target AI nominates, regardless of the model’s confidence or how high it ranked, arrives at a bench where someone has to determine whether it’s actually druggable. It’s the same bench work that drug discovery has always run on, now with a longer list to work through.
After the Algorithm Runs
The algorithm ranks targets by likelihood of success. It has nothing to say about how to run the first experiment on something nobody has studied before. That part is up to the researcher, and for AI-nominated targets, it starts from nothing: no established pharmacology, no known ligands, no prior assay to reach for. The biology arrives before the playbook. The researcher is the one who writes it.
Researchers spend most of their careers in that work. A breakthrough is one possible outcome, but it isn’t the only one that matters. The ordinary work never stops, and nothing happens without it.
What makes that work possible has always been the same: the questions a program can ask are bounded by what its assays can actually measure. For AI-nominated targets, that boundary matters more than ever. The algorithm predicts. The cell decides. NanoBRET™ Target Engagement was built to answer exactly that question: not whether a compound binds in a test tube, but whether it engages the target inside a living cell, in the environment where the biology is actually happening. The tools are what make validation possible, and validation is where hypotheses become evidence.
Nobody’s Filming This
Did you see the keynote about the researcher who opened a notebook to design the first experiment on a target the algorithm was most confident about, and found nothing in the literature to work from? Or the press release about remaking the buffer at eleven o’clock on a Tuesday because the result still doesn’t make sense? Or the headline about navigating three approval screens to order the compound the algorithm ranked first, waiting on a backorder notice, finding an alternative supplier, and resubmitting, because the most promising candidate in the dataset is temporarily out of stock?
No. Because the ordinary day-to-day doesn’t stand a chance next to the algorithm.
Somewhere between the algorithm and the answer, there’s a bench. Nobody’s filming this, but everything else depends on it.
References
Pun F.W. et al. (2026) Target identification and assessment in the era of AI. Nat. Rev. Drug Discov. https://doi.org/10.1038/s41573-026-01412-8
Elise Johnson
Latest posts by Elise Johnson (see all)
- AI Identifies the Target. Someone Still Has to Validate It. - July 9, 2026
- Another Crack in the RAS Code: The Measurement Story Behind Daraxonrasib - June 25, 2026
- What Counts as Evidence - May 7, 2026
