WASHINGTON - Harvard researchers have discovered half a dozen new genes involved in autism that suggest the disorder strikes in a brain that can't properly form new connections.

The findings also may help explain why intense education programs do help some autistic children -- because certain genes that respond to experience weren't missing, they were just stuck in the "off" position.

"The circuits are there but you have to give it an extra push," said Dr. Gary Goldstein of the Kennedy Krieger Institute in Baltimore, which wasn't involved in the gene hunt but is well-known for its autism behavioural therapy.

The genetics suggest that "what we're doing makes sense when we work with these little kids -- and work and work and work -- and suddenly get through," he said.

But the study's bigger message is that autism is too strikingly individual to envision an easy gene test for it. Instead, patients are turning out to have a wide variety, almost a custom set, of gene defects.

"Almost every kid with autism has their own particular cause of it," said Dr. Christopher Walsh, chief of genetics at Children's Hospital Boston, who led the research published in Friday's edition of the journal Science.

Autism spectrum disorders include a range of poorly understood brain conditions, from the mild Asperger's syndrome to more severe autism characterized by poor social interaction, impaired communication and repetitious behaviours.

It's clear that genes play a big role in autism, from studies of twins and families with multiple affected children. But so far, the genetic cause is known for only about 15 per cent of autism cases, Walsh said.

So Walsh's team took a new tack. They turned to the Middle East, a part of the world with large families and a tendency for cousins to marry, characteristics that increase the odds of finding rare genes. They recruited 88 families with cousin marriages and a high incidence of autism, from Jordan, Saudi Arabia, Kuwait, Oman, Pakistan, Qatar, Turkey and the United Arab Emirates. They compared the DNA of family members to search for what are called recessive mutations -- where mom and dad can be healthy carriers of a gene defect but a child who inherits that defect from both parents gets sick.

In some of the families, they found large chunks of missing DNA regions that followed that recessive rule. The missing regions varied among families, but they affected at least six genes that play a role in autism.

Here's why this matters: All the genes seem to be part of a network involved in a basic foundation of learning -- how neurons respond to new experiences by forming connections between each other, called synapses.

In the first year or two of life -- when autism symptoms appear -- synapses rapidly form and mature, and unnecessary ones are "pruned" back. In other words, a baby's brain is literally being shaped by its first experiences so that it is structurally able to perform learning and other functions of later life.

"This paper points to problems specifically in the way that experience sculpts the developing brain," explained Dr. Thomas Insel, director of the National Institute of Mental Health, which helped fund the work.

Some earlier research had pointed to the same underlying problem, so these newly found genes "join a growing list to suggest that autism is a synaptic disorder," he said.

If that sounds discouraging, here's the good news: The missing DNA didn't always translate into missing genes. Instead what usually was missing were the on/off switches for these autism-related genes. Essentially, some genes were asleep instead of doing their synapse work.

"I find that hopeful" because "there are ways that are being discovered to activate genes," Walsh said. "This might be an unanticipated way of developing therapies in the long term for autism: Identifying these kids where all the right genes are present, just not turned on in the right way."

At Kennedy Krieger, Goldstein thinks the work may provide a gene-level explanation for why some children already are helped by intense therapy.

"We have trouble getting through to these children, but with repeated stimulation we can do it," he said. "These are circuits that have an ability not so much to recover but to work around the problem."