The US Government is beginning a project designed to unlock the genetic abnormalities that contribute to cancer, an effort that would exceed the Human Genome Project in complexity and could eventually lead to new diagnostic tests and treatments for the disease, said a report.

The Government would spend $100 million over three years on a pilot phase of the project, which will be called The Cancer Genome Atlas. The cancer institute will contribute half of the money with the other half provided by the National Human Genome Research Institute. Both are part of the National Institutes of Health.

Scientists have long known that genetic mutations that accumulate in normal cells over a person's lifetime can make those cells cancerous.

About 300 genes involved in cancer are already known, and there are a handful of drugs that work by interfering with specific genetic abnormalities.

Federal officials and many cancer researchers say that a more systematic search could find many more genes and gene variations that play an important role in determining how aggressive a cancer will be and what drugs might work best. The first fruits, such as new diagnostic tests, may be seen in several years, the report says.

The project would involve determining the sequence of letters in the DNA of tumour cells obtained from biopsies or surgery. The initials of the project's name, T.C.G.A., represent the four letters of the genetic code.

Scientists will also look for other changes like duplications or deletions of genes, or differences between cells in which genes are turned on or turned off.

In February a committee advising the National Cancer Institute proposed such a project, which it estimated would cost $1.35 billion over nine years.

The results of the three-year pilot project will be assessed first before deciding whether a full program would be worthwhile. The pilot phase will involve studying hundreds of tumour samples from two or three types of cancer, which have not been chosen yet. The decision will depend upon factors like the availability of tumour samples. A full project could involve studying 50 types of cancer.

Some scientists have expressed concern about the project, saying it may divert grant money from individual investigators at a time when the N.I.H. budget is not growing rapidly, as it was a few years ago.

Some have also said that cancer cells are so genetically heterogeneous - two cells in the same tumour can differ in their mutations - that it will be hard to find mutations that are meaningful and could lead to new drugs.

To minimize the risk of failure, the pilot project will choose types of cancer that have relatively little variability, officials said.

The amount of DNA sequencing involved could easily exceed what was done in the Human Genome Project, which determined the sequence of the three billion DNA units in human chromosomes. Since each cancer cell contains a complete genome, determining the full sequence of thousands of cells will be like doing thousands of genome projects.

The findings will be put in one or more databanks and be available free to researchers, officials said. They said they expected that medical foundations, companies and academic researchers would take part in the project, the report said.