While I’ve been thinking about the reproducibility of ‘omics research and conflict of interest at Duke (more on that later), a lot of cool cancer-related science papers have been published. The latest one to catch my eye, published online today in Nature, suggests that a four-gene panel may help identify dangerous prostate cancer cases.
When tested on three independent data sets, the four-gene signature had nearly the same prognostic value as Gleason score and significantly improved the prognostic accuracy when used in combination with either Gleason score.
For example, when the investigators used immunohistochemical staining to measure expression of their four genes on 405 tissue microarray samples from the Physicians’ Health Study (38 lethal cases and 367 indolent cases), they found the prognostic accuracy of the four-genes was C=0.829, compared with C=0.774 for Gleason score.
A model with all clinical-pathologic features, including Gleason score, age at diagnosis, and TNM stage, had a prognostic value of C=0.842, which increased to 0.913 when the four-gene signature was added in.
A similar trend was seen in two other data sets when the investigators, led by
Ronald A. DePinho, MD, Director of the Belfer Institute for Applied Cancer Science at Dana-Farber Cancer Institute, compared their gene signature with Gleason score alone and with the signature and Gleason score together.
Unfortunately, however, they did not report whether the gene signature was able to add any prognostic accuracy to the more inclusive clinical-pathological model in these other datasets. And after all, that is really the test to beat, not just Gleason score on it’s own.
That criticism aside, the paper is interesting because of how the team derived the four-gene signature.
Instead of looking for gene expression that correlates with disease progression, the Boston group started with a mouse that lacks PTEN expression in the prostate and develops prostatic intraepithelial neoplasia.
They then looked for gene expression differences between these lesions and normal prostate tissue and saw that the TGFb/SMAD4 pathway was upregulated.
Because SMAD4 is often downregulated in human prostate tumors, the investigators reasoned that high SMAD4 expression was keeping abnormal growth in check in the PTEN-deficient prostate.
To find out, they knocked out SMAD4 expression in the prostate of PTEN-deficient animals and saw that more aggressive tumors formed. They subsequently found that cyclin D1 was critical for increasing tumor cell proliferation in the animal model and that SPP1 increased metastasis.
And thus, they came up with their four gene signature: PTEN, SMAD4, cyclin D1, and SPP1.
“We wanted causal genes,” Dr. DePinho told me. “We reasoned they would be more predictive. Other signatures use a guilt-by-association approach, which are just surrogates for what is going on.”
Whether this four-gene signature will improve care for men with prostate cancer is yet to be seen, but Dr. DePinho and his colleagues think it will. They have already patented the test and licensed it to a company for commercial development and marketing as early as 2012. (Dr. DePinho states openly that he has a financial conflict of interest in the project.)