In silico drug modelling is the use of computer models of the human body to predict how a medicine will behave in people, before it is tested in people. In silico means "done on a computer". The models predict three things: how much of the medicine is absorbed into the blood, how it acts on the body, and how fast the body removes it. It is also called in silico pharmacology, or pharmacokinetic modelling.
Scientists describe what the body does to a medicine with four letters, ADME. A is absorption: how much of the medicine enters the blood from the gut or the skin. D is distribution: which organs it reaches. M is metabolism: how the liver breaks it down. E is excretion: how the kidneys or the gut remove it. What the medicine does to the body, such as lowering blood pressure, is studied separately.
A common kind of model is the PBPK model, short for physiologically based pharmacokinetic model. It represents the body as a set of organs, each with its own size and blood supply. The scientist enters the molecule's measured properties, such as how well it dissolves and how fast liver cells break it down. The model then calculates the level of the medicine in the blood, hour by hour.
Say a pharmacology team at a drug company in Chennai has a new tablet for high blood pressure. Liver-cell tests show that the molecule is broken down quickly. The team builds a PBPK model of an average adult. The model predicts that a 50 mg tablet would reach its highest blood level after 2 hours, and that half of it would be gone after 4 hours. So a single daily tablet would not keep the blood pressure down at night. The model shows that a 100 mg slow-release tablet, taken once a day, would keep the level above the useful minimum for 24 hours. The team tests that design first, instead of trying several doses in volunteers.
The same models answer other questions. What dose suits a child, whose liver is smaller? What happens if the patient also takes another medicine that slows the liver down? Drug regulators, including the US FDA, accept such model results in some applications, for example to show that two medicines can safely be taken together.
The molecules usually arrive from AI drug discovery or from molecular modelling software. The predictions then help plan the first trial in people. Clinical trial design software uses them to choose the doses and the number of patients. The trial itself is described in clinical trials.
How is this different from molecular modelling? Molecular modelling predicts how one molecule fits one protein. In silico drug modelling predicts what happens to the molecule in a whole body. Neither replaces testing. The models only decide which tests are worth doing first. The data they rely on comes from the laboratory, much of it described in genomics and the research laboratory.