There was a time when a laboratory experiment began with a scientist deciding what to try.
Increasingly, it may begin with a computer suggesting what is worth trying.
A sequence is analysed. A protein is modelled. Thousands of biological possibilities are compared. A system identifies a promising target, predicts how a molecule might behave or suggests which experiment is most likely to produce useful information.
Then the scientists go back to the laboratory.
This does not mean computers have replaced biology. The physical world remains stubbornly complicated, and predictions still have to survive contact with cells, tissues and living organisms.
But something important is changing.
Bioinformatics, once largely associated with organising and analysing biological data, is becoming a foundation for modelling biology itself.
That shift could prove much bigger than the name of the discipline suggests.
The new bioinformatics is less about analysing data and more about modelling life
Biology produces extraordinary quantities of data. DNA sequences, RNA, proteins, cellular measurements, imaging, clinical records and countless other signals can now be collected at scales that would have been difficult to imagine not long ago.
The challenge is no longer simply obtaining information.
It is understanding how the pieces relate to one another.
A gene does not operate in isolation. A protein does not exist independently of the cellular environment around it. A disease can emerge from the interaction of genetics, cell state, tissue conditions, age and other factors.
That is where modern computational biology is beginning to move beyond traditional analysis.
Researchers are developing large biological foundation models that can learn patterns from genomic and other biological datasets and then be adapted to different tasks. Recent work has shown models trained across large functional-genomics datasets performing multiple forms of genome interpretation, while other foundation models are integrating information about three-dimensional genome organisation, chromatin activity and cellular state. ([nature.com](https://www.nature.com/articles/s41467-026-73129-6?utm_source=chatgpt.com))
This is an important change in philosophy.
Instead of building one computational tool for one biological question, researchers are increasingly asking whether a sufficiently capable model could learn a broader language of biology and then be used across many questions.
That is the same conceptual shift that has transformed other areas of computing.
The result is a new possibility: biology becoming increasingly computable.
Once biology can be modelled, experimentation can change
For centuries, biological discovery has depended heavily on experimentation.
Form a hypothesis. Run an experiment. Observe the result. Adjust the hypothesis. Try again.
It is a powerful process, but it can also be painfully slow.
Living systems are expensive to manipulate. Some experiments take weeks or months. Others require specialised equipment, scarce biological material or complex animal and clinical studies.
A computational model offers another layer.
Researchers can explore large numbers of possibilities before selecting a smaller number for physical testing.
This does not eliminate the laboratory. It changes what the laboratory is asked to do.
Instead of searching blindly through thousands of possibilities, researchers may increasingly use computation to narrow the space first.
This is already emerging in areas such as molecular and drug discovery. Recent reviews describe AI moving from sequence analysis and structure prediction toward generative approaches that can propose new biological molecules and help optimise candidate therapies. ([pubmed.ncbi.nlm.nih.gov](https://pubmed.ncbi.nlm.nih.gov/42453397/?utm_source=chatgpt.com))
The commercial implication is significant.
A company that can reduce the number of experiments required to identify a promising candidate is not merely selling software.
It may be changing the economics of research.
Every failed experiment that can be ruled out computationally represents time. Every promising molecule identified earlier represents a potential advantage. Every biological interaction that can be understood before entering a laboratory creates another opportunity to compress the development cycle.
This is one reason the boundary between bioinformatics, AI and biotechnology is becoming less distinct.
The companies that once sat in separate industries are beginning to occupy the same computational territory.
The next major biological resource may be a model rather than a database
For years, one of the great assets in life sciences has been data.
Genomic databases, clinical datasets, molecular libraries and collections of biological measurements have become enormously valuable.
But raw data has a limitation.
You still need someone or something capable of interpreting it.
A sufficiently capable biological model could change the economics of that process.
Instead of repeatedly querying large datasets with separate tools, researchers could interact with systems that have learned broad patterns across multiple biological domains.
This is one reason the rise of biological foundation models is attracting so much attention. A recent survey of the field describes a rapidly expanding ecosystem of models trained on different biological modalities and designed for tasks ranging from molecular prediction to cellular and genomic analysis. ([nature.com](https://www.nature.com/articles/s41587-026-03135-y?utm_source=chatgpt.com))
The ambition is becoming even broader.
Researchers are beginning to discuss computational representations of organisms that combine information across biological scales, from molecules and cells to tissues and individuals. The vision is not a perfect digital copy of a human being. It is a computational framework that can simulate enough of biological complexity to make experiments more targeted and decisions more informed. ([nature.com](https://www.nature.com/articles/s41591-026-04595-0?utm_source=chatgpt.com))
That is a profound shift in what a biological company could eventually own.
The valuable asset may no longer be only a database of information.
It may be the model that understands relationships within that information.
And once the model becomes useful enough, access to it could become a platform in its own right.
AI may turn bioinformatics into an industry of biological infrastructure
This is where bioinformatics becomes especially interesting from a business perspective.
The field has traditionally lived somewhat behind the scenes. It provides analysis, pipelines, databases and computational tools that enable researchers to do other things.
That role may be changing.
As biological models become more capable, the software layer can move closer to the centre of the research process.
A modern biotechnology company may need systems that can interpret genomic variation, analyse single-cell data, predict molecular behaviour, identify promising therapeutic targets, compare experimental results and help researchers decide what to test next.
Those functions may increasingly share a common computational foundation.
That creates an ecosystem opportunity.
Some businesses will build biological foundation models.
Others will provide training data, specialised datasets or computational infrastructure.
Some will build tools for laboratories to interact with the models. Others will specialise in particular areas such as genomics, proteins, cancer, drug discovery or synthetic biology.
Still others may build the validation systems needed to determine whether an AI-generated biological prediction is actually reliable.
There is already evidence that major technology companies are moving further into life sciences and computational biology, while new evaluation efforts are emerging to determine whether AI systems can perform complex bioinformatics tasks reliably. ([nature.com](https://www.nature.com/articles/s41587-026-03271-5?utm_source=chatgpt.com))
This is an important signal.
When technology companies begin entering a scientific domain, they do not necessarily want to become traditional laboratories.
They often want to build the infrastructure around the laboratories.
Bioinformatics may be moving toward precisely that kind of infrastructure role.
The most valuable biological AI may still need a human expert
There is a temptation to think that increasingly capable AI will eventually make bioinformaticians unnecessary.
The more interesting possibility is almost the opposite.
The value of expertise may rise because the volume of machine-generated possibilities becomes so large.
An AI system can identify a pattern. It does not automatically know whether that pattern reflects meaningful biology, biased data, an experimental artefact or an attractive coincidence.
Researchers working on AI and bioinformatics increasingly emphasise that human expertise remains essential for data curation, model design, interpretation and scientific validation. A recent perspective on the changing role of bioinformatics argues that AI should be seen primarily as an accelerator whose value depends on expert guidance rather than as an autonomous replacement for biological judgement. ([nature.com](https://www.nature.com/articles/s41746-026-02777-1?utm_source=chatgpt.com))
This may create a new kind of scientist.
Someone who understands biology deeply enough to ask good questions, computation well enough to understand what the models are doing, and experimental science well enough to determine whether the answer is plausible.
The winning teams may therefore not be the ones with the largest AI system alone.
They may be the teams capable of creating a continuous loop between computation and experiment.
The model proposes.
The laboratory tests.
The result returns to the model.
The model learns.
The next experiment becomes better informed.
That loop could become one of the defining structures of modern biological research.
The long-term prize is not faster analysis. It is programmable biology
This is where the field could become truly transformative.
Understanding biology is one achievement.
Predicting biology is another.
Designing biology is something else entirely.
The movement from analysis toward prediction and design is already visible in areas such as protein and RNA engineering, where generative AI is being explored to propose sequences and structures rather than merely analyse molecules that already exist. Recent work describes the increasing use of generative models for RNA design and biological sequence engineering, although experimental validation remains essential. ([nature.com](https://www.nature.com/articles/s41576-026-00967-x?utm_source=chatgpt.com))
If this trajectory continues, bioinformatics could become part of an entirely new industrial model.
Instead of finding a biological molecule and then trying to make it useful, researchers could increasingly describe the function they want and computational systems could propose biological candidates worth testing.
Instead of discovering every useful biological process through trial and error, models could help researchers explore the possibilities first.
Instead of treating biological complexity as something that must simply be measured, researchers could begin treating parts of it as something that can be modelled and engineered.
That is a very different relationship with biology.
It does not mean that life has become predictable. It has not.
Biology will continue to surprise us, and the distance between a computational prediction and a working biological intervention can still be enormous.
But the direction is unmistakable.
The computer is becoming part of the biological laboratory’s imagination.
And that may be the most important thing happening in bioinformatics right now.
The discipline began by helping scientists handle biological information.
It may be heading toward something much larger: becoming the computational layer through which humanity learns to read biology, reason about it and eventually design parts of it.
The companies that build that layer will not simply make research faster.
They may help determine what biology can become.