2026: the year AI stops being optional in drug discovery

Artificial intelligence (AI) is moving from isolated applications into the core of drug discovery. By 2026, it is expected to influence how targets are identified, how biological data are analysed and how clinical development decisions are made.

As AI tools are applied earlier in discovery, computational prediction is increasingly used alongside experimental validation rather than after it. This allows teams to test biological hypotheses earlier, build confidence sooner and reduce late-stage failure.

AI-guided target identification becomes the starting point

In early discovery, target identification has traditionally involved reviewing published evidence, developing biological hypotheses and validating them through repeated laboratory experiments. This process can be slow and often relies on partial views of disease biology. By 2026, early target selection is expected to depend far more on computational analysis, enabling scientists to interrogate large biological datasets before committing to wet-lab work.

“In 2026, identifying disease targets will rely on in silico exploration before any wet-lab validation begins,” says Veronica DeFelice, Director of Biologics at Sapio Sciences. She explains that AI-guided platforms connected to laboratory information management systems will integrate genomic, proteomic and transcriptomic datasets to reveal molecular patterns and disease mechanisms that were previously hidden when data were analysed in isolation.

By combining these datasets, scientists can define more precise starting points for biologics discovery. DeFelice notes that researchers will use these insights to select targets “with stronger biological rationale and a clearer therapeutic pathway based on disease association and molecular verification.” This, she says, will reduce the number of programmes that stall during preclinical development. Earlier confidence in target biology also allows teams to focus resources on candidates with a clearer line of sight to mechanism, modality selection and downstream development decisions, rather than revisiting foundational assumptions later in the pipeline.

Biological modelling moves into everyday workflows

Alongside AI-guided target identification, biological modelling is set to become foundational in early discovery rather than a specialist activity. DeFelice points to the growing availability of structure prediction and binding simulation tools embedded directly within AI-native laboratory systems and digital lab notebooks.

By 2026, scientists working on complex biologic modalities such as multispecific antibodies and fusion proteins will routinely evaluate affinity and specificity computationally before committing resources to experimental work. Making modelling part of daily scientific practice reduces trial and error and improves candidate selection at earlier stages.

This closer integration of computational prediction and experimental validation is changing how discovery teams work in practice. Rather than operating in sequence, modelling and wet-lab experiments increasingly inform one another, allowing data generated in one step to guide the next. This shortens feedback cycles and helps teams move from biological insight to candidate selection more efficiently.

Genomics data reaches unprecedented scale

The growing reliance on AI in discovery is closely linked to the scale of biological data now being generated, particularly in genomics. Advances in sequencing technologies continue to accelerate data generation, but analysis remains a major bottleneck.

“In 2026, we will see stunning examples of how AI can interpret genomics data to understand complex biology,” says Neil Ward, Vice President and General Manager for EMEA at PacBio. He highlights the scale of the challenge, noting that analysing a single genome involves tens of thousands of lines of code. At population scale, genomics studies “could generate up to 15× more data than YouTube over the next decade.”

At this scale, traditional bioinformatics approaches alone are no longer sufficient. Ward points to a growing move towards natural language interfaces that allow scientists to interrogate genomics data without relying exclusively on specialist code.

“We are already seeing major AI players partnering with science firms to analyse genomics data in natural language rather than relying on specialised bioinformatics code,” he says. These collaborations are intended to make complex genomic analysis easier to use without removing the underlying analytical complexity.

Digital twins move from pilots to practice

AI is also increasingly being used in later stages of development, particularly through digital twins. According to Dr Gen Li, Founder and President at Phesi, 2026 will mark a turning point for their adoption.

“After years of experimentation, 2026 will mark the year digital twins move from pilot to practice in clinical development,” Li says. Sponsors have increasingly explored digital twins to optimise protocol design, reduce amendments and accelerate timelines, but uncertainty around regulation has limited wider use.

“That’s now changing,” Li explains. Regulators including the FDA are expanding AI frameworks and finalising risk-based guidance to support safe and effective use of these technologies in clinical development. As regulatory clarity improves, opportunities to integrate digital twins into trial design and execution are expected to grow.

Li emphasises that regulatory trust is central to adoption. “To unlock the full value of digital twins, sponsors must earn regulatory trust through rigorous validation, ethical data governance and clear documentation,” he says. Continued collaboration between regulators, sponsors and technology partners will be essential.

If implemented responsibly, digital twins offer the potential for faster, more patient-centric and more equitable clinical trials.

Towards a more integrated discovery process

Taken together, these developments point to a more integrated and computationally driven approach to drug discovery and development. AI-guided target identification, embedded biological modelling, scalable genomics analysis and digital twins are increasingly used together rather than as isolated tools. As AI becomes more widely embedded, discovery teams are moving from experimentation to routine use, with the focus now on where these tools sit in existing workflows and how their outputs inform early scientific decisions.