By Bill Berkrot

(Reuters) - U.S. researchers on Thursday said they had found a way to predict male sexual orientation based on molecular markers that control DNA function, but genetics experts warned that the research has important limitations and will not provide definitive answers to a potential biological basis for sexual preference.

Findings from the study, which has yet to be published or reviewed in detail by other scientists, were presented at a meeting of the American Society of Human Genetics in Baltimore.

It followed 37 pairs of identical male twins in which one was homosexual and one heterosexual, and 10 sets of twins in which both males were homosexual. The study found that the presence of specific epigenetic marks in nine areas of the human genome could predict homosexual preference with up to 70% accuracy.

The epigenome is sometimes described as molecular "switches" that can turn on or silence individual genes in DNA. Scientists believe epigenetic differences can be influenced by environmental and lifestyle factors, from exposure to chemicals to parental nurturing.

"To our knowledge, this is the first example of a predictive model for sexual orientation based on molecular markers," Tuck Ngun, lead researcher on the study from the David Geffen School of Medicine of the University of California, Los Angeles, said in a statement.

Genetics experts who critiqued the findings said it was premature to draw any conclusions on the predictive powers of epigenetic markers.

"The question as to whether that prediction is going to be useful outside of the small number of twins in the study is really unclear," said Dr. Christopher Gregg, a genetics professor at the University of Utah.

Others noted the small size of the population studied and stressed that such findings often fall apart when applied to larger groups of people.

"One thing you can clearly see is that the sample size is too small. They don't have enough power to make that claim," said Dr. Peng Jin, professor of human genetics from Emory University in Atlanta, who attended the meeting in Baltimore.

"What they are seeing may be certain correlations, but I don't think they have what they claim, which is a predicting model," he added. "It's definitely an interesting observation, but ... I don't want the general audience to misinterpret whatever they are presenting," Jin said.

Gregg said he was impressed by the UCLA team's "state-of-the-art" methodology, but said much larger studies must be undertaken to reach any conclusions.

"Just because there is something different doesn't mean that's what's causing people to behave one way versus the other," he said.