Before long, we will each have an alter ego to assess the medication we need. That's the vision of Natalia Alexandrov, winner of the New Scientist/NC3Rs "Beyond animal research" essay competition.

A QUIET night in December 2050. Small town anywhere on Earth. In the hospital, little Peter has just made his appearance in the world. As the happy family takes pictures of mother and baby, the thought that his birth is incomplete is far from their minds. And the birth will not be complete for a few hours, until hundreds of miles away his virtual twin is born.

While Peter is dreaming his first dreams, samples of his blood and tissue are analysed and the resulting data transmitted to the simulation and modelling department of the regional medical centre.

It is expensive and time-consuming to build a virtual human from scratch, so a library of averaged mathematical models of newborns is maintained. The initial models differ by sex, ethnicity, geographic origin, basic genetic make-up and other salient characteristics. When Peter's data arrive, they are integrated with a model whose attributes most closely resemble Peter's. The vast resulting computational model is as unique as Peter himself.

Throughout Peter's childhood his parents take him for medical check-ups, inoculations and treatment for a broken arm and colds. All this information makes his virtual twin grow. When the boy is 10, flu leaves him with a complication - severe bronchitis. Which antibiotic to prescribe? The family doctor downloads Peter's virtual twin, updates it with the latest tests, and runs simulations for the range of available antibiotics to anticipate short-term and long-term effects. This identifies both the perfect drug and one that would have had a life-threatening, long-term effect on Peter's blood-clotting ability, possibly leading to a future stroke.

Why an individual twin? The laboratory animals used to develop drugs do not always represent humans adequately, and even well-tested drugs can cause adverse reactions in some people.

How far are we from building useful and practical virtual twins? There are many sophisticated models of individual organs and systems, such as network models of the metabolic, immune, nervous and circulatory systems; computational fluid dynamics models of blood flow; structural models of the heart and other muscles. These models will continue to grow in fidelity, but the behaviour of a complex system is a function of its complexity, and the biggest barrier is, arguably, the integration of subsystem models into one systemic model.

As well as the difficulty of modelling every aspect of the organism in terms amenable to computation, the challenges are in combining disparate models and the sheer volume and complexity of the information needed. However, in this writer's opinion, the situation is by no means hopeless. Moore's law of growing computational resources will provide the necessary storage and speed. The complex mathematics involved will be more challenging, but by no means prohibitively so.

The benefits of a virtual twin that evolves along with the human it represents would be enormous. The action and long-term consequences of most drugs are at present a mystery, despite careful research and animal testing. We cannot quantify the processes actually taking place in our bodies as a result of an illness or in response to a drug. A faithful computational representation of an individual would yield quantifiable, rigorous measurements of the model's reactions. There would no longer be a need to identify a population for a drug trial, nor would years go by in tests while desperate patients are waiting for a new drug. A set of virtual patients with precisely known characteristics could be selected at any time, and drug research conducted at computational speeds, resulting in safe, effective treatments based on individual patients' characteristics.