Vini Jain, content and research team partner at Compound Vitality, explains how certain advancements in the preclinical stage are improving the drug discovery space.
In July 2026, Insilico Medicine started a Phase III trial of rentosertib for idiopathic pulmonary fibrosis in China. On its own, that’s routine pipeline news. What makes it worth a second look is where the drug came from. Its target, a kinase called TNIK, and the molecule that blocks it were both found with generative AI, and a year earlier the TNIK inhibitor had produced the first randomised Phase IIa results for a drug discovered that way.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
It’s tempting to read stories like this as proof that discovery is simply getting faster. Some of it is. The more useful observation is that the hard parts are moving. Computational design has made the front of the pipeline quicker and cheaper, regulators are rewriting what counts as preclinical evidence, and early trials are being asked to carry more of the science. Sponsors who plan around where the friction now sits will get far more out of these tools.
Design is cheap now. Making and testing isn’t
Peptides show the shift clearly. The success of glucagon-like peptide-1 receptor agonists (GLP-1RA) drugs has put the modality back at the centre of many pipelines, and design tools have arrived to meet the demand. In 2025, researchers at the University of Washington described RFpeptides, a deep-learning pipeline for designing protein-binding macrocycles from scratch. They tested 20 or fewer designs against each of four protein targets and found binders for all four. Until recently, getting that far usually meant large, expensive library screens.
Every one of those designs still has to become a physical molecule before anyone can test it. In practice, that means synthesising or sourcing small lots of research peptides, confirming identity and purity, and running binding assays, often over several rounds. Generating candidates is now fast. Synthesis and assay work hasn’t kept pace, so the wet lab increasingly sets the tempo of early discovery.
It also raises the stakes on that early material. When a model proposes twenty candidates instead of a screen sifting through millions, each assay result carries more weight. A misleading one, from an impure sample or a salt form that behaves differently from the intended molecule, can send a programme down the wrong path quickly. Binding is also only the first filter. A macrocycle that grips its target in a dish still has to survive proteases, reach the right tissue and stay in circulation long enough to matter, and those properties remain far harder to predict. Fast design rewards careful chemistry.
Preclinical evidence is being rebuilt around human biology
The next stage is changing for different reasons. The legal groundwork came first: the FDA Modernization Act 2.0, signed in late 2022, removed the statutory requirement to test drugs in animals before human trials. In April 2025, the FDA set out a roadmap for reducing animal testing in preclinical safety studies, beginning with monoclonal antibodies. A draft guidance in December recommended reducing or dropping long-term primate toxicity studies for certain antibodies, and in April 2026 the agency said it had met its first-year goals. The tools it wants sponsors to lean on are new approach methodologies: organoids, organ-on-chip systems, cell-based assays and computational models.
For now, this changes the conversation more than the workload. Peptides and small molecules aren’t the first focus, and I wouldn’t plan an IND around skipping animal studies just yet. But the logic is shifting. A preclinical package built on human-relevant data, with animal work kept for the questions it genuinely answers, is clearly where things are heading. Teams that start generating that data now will have something to show when the scope widens.
Early trials now carry more of the science
Better discovery also changes what a sponsor knows on the way into the clinic. A candidate that came from a defined computational hypothesis arrives with a specific mechanism attached, and that invites early trials to test the mechanism alongside safety. Rentosertib’s Phase IIa is a good example. As well as lung function, the team measured serum proteins and reported dose-dependent falls in profibrotic markers and rises in anti-inflammatory ones that tracked with the clinical changes.
That’s valuable, and it isn’t free. Biomarker-heavy early studies need assays validated before the first patient is dosed, samples handled consistently across sites, and protocols with room for the questions a novel target raises. For trial teams, the practical effect of better discovery is often more work in Phase I and IIa, traded for clearer go or no-go decisions later. It also changes who needs to be involved early. Translational scientists, bioanalytical labs and clinical operations increasingly have to plan Phase I together instead of handing work along in sequence.
What hasn’t changed
None of this has repealed attrition. A 2025 benchmarking study of R&D success rates at 18 leading pharmaceutical companies found an average likelihood of first approval of 14.3%, measured from Phase I. A separate analysis in Nature Communications found that clinical trial success rates declined through the early 2000s, plateaued, and have only recently begun to rise. Better discovery may be part of that recovery. It hasn’t rewritten the odds.
Phase II is still where most programmes stall, because it’s the first real test of whether a mechanism matters in patients, and no model fully answers that in advance. Manufacturing doesn’t get easier either. A molecule designed in days still needs a scalable synthesis, a controlled impurity profile and GMP supply before it reaches a trial.
Planning for the new bottlenecks
The practical lesson is to put resources where the friction has moved. Invest in assay capacity and material quality early, since that’s where fast design gets checked. Talk to regulators about non-animal data before the preclinical package is locked. And design early trials that can confirm or kill a mechanism quickly, so Phase II isn’t the first time anyone finds out.
Rentosertib still had to clear every stage any other drug does. What changed was how it reached the starting line and how much its developers understood when they got there. That, more than raw speed, is the real change in the path from research to clinical trials, and it’s where sponsors should be paying attention.
