LOS ANGELES - The old joke is that of fast, cheap and good, you can have any two.

But Francis Collins, director of the National Institute for Human Genome Research, says that's really no longer true in gene sequencing.

From 2002 to 2007, the cost of a sequencing study to find a disease-associated gene has gone down by four orders of magnitude, he said at the opening plenary session of the American Association for Cancer Research annual meeting. That drop, from $10 billion to $800,000, is due to a combination of sequencing advances and the selection of a subset of sentinel single-nucleotide polymorphisms (SNPs) from the human genome.

That technological advance has allowed the study of cancer genetics to improve in several ways. For one thing, it has allowed scientists to search the entire genome for genes associated with cancer, rather than studying only those genes that appear likely candidates already. As Collins put it, selecting certain genes to test "limits us to the hypotheses that we can come up with - and we can all agree that there's a lot we don't know" about cancer genomics.

The downside of large association studies is that the larger the number of SNPs one studies, the greater the chance of false positives. But sheer sample sizes can make up for that. Collins said that a study now under way on the association of certain SNPs with breast cancer has uncovered several candidates whose P-value is less than 10-9, with the lowest p value being less than 10-74. Collins noted dryly, "You'd probably have to believe that."

While Collins lectured mainly on the somatic mutations that lead to cancer, Mary Claire King from the University of Washington spoke on inherited mutations that confer cancer risk. But she, too, noted the enormous genetic heterogeneity of cancer. "Every family with cancer is unhappy in its own way," she told the audience.

King, whose lab nailed down the association between the BRCA genes and cancer risk, described another type of association that the new large-scale studies are better at detecting: weak associations - that is, mutations in genes that lead not to the extremely high risk the BRCA mutations confer, but a more moderate doubling of cancer risk. Such associations have been harder to untangle because they need large numbers of subjects to convincingly demonstrate an association.

King discussed her work on genes with such weaker associations at the plenary session, but also noted the substantial challenges that those genes will bring to the clinic once they are discovered. BRCA mutations confer an extremely high lifetime risk of developing cancer - with estimates reaching up to 85 percent for some BRCA1 mutations - and so even radical measures such as preventive mastectomies seem reasonable to at least consider.

On the other hand, a gene that doubles risk leads to a 25 percent lifetime risk of developing breast cancer. She said that "as scientists, we are prepared to live with this uncertainty," but added that it's more difficult for a clinician who must translate such risk assessments into recommendations for what someone with a mutation should actually do. "We are working very hard to provide you with answers for your patients."

She also noted that if many genes conferring comparatively slight risk are indeed identified, it will become enormously important to provide women with access to diagnostic screening so that knowledge can be used for effective prevention, which is not a cheap proposal. King called the case for single-payer insurance "compelling" in that context.

That's not something insurers will be happy to hear, especially when followed by the advice of Charles Sawyers, from the Memorial Sloan-Kettering Cancer Center. "We must move quickly to rational combination therapy," he told the audience.

Sawyers described two different types of combination therapies. One is the simultaneous use of several drugs against sequential mutations in a single target to avoid the development of resistance. An example is the use of imatinib and dasatinib, both of which block a hyperactive bcr/abl kinase, to treat chronic myelogenous leukemia. Currently, imatinib and dasatinib are used sequentially, but Sawyers said he believes that strategy needs to be "reconsidered, and potentially replaced" by combination therapy.

The second, even more challenging strategy is to inhibit multiple downstream pathways that are all activated by the loss of a tumor suppressor. He used the example of PTEN loss, which activates a number of pathways, including the mTOR and jun kinase pathways.

Sawyers noted that the design of such trials has major challenges, from picking the type of cancer to study - because the loss of a tumor suppressor can cause many different types of cancers - to enrolling subjects. In one trial, to test the use of rapamycin against mTOR in glioblastoma, only 15 of 154 patients ultimately met the criteria to be treated, so patient accrual for sizable trials undoubtedly will be tough.

Still, the strategy paid off: 50 percent of patients in the study responded, as opposed to a response rate that would have probably been less than 20 percent if the study had treated all comers with rapamycin. Sawyers called such patient enrichment the "only intelligent way" to move forward; otherwise, disappointing response rates will lead drug developers to jettison good drug candidates.

The conference continues through Wednesday.