High-Tech Detectives Profile Gene Expression to Tailor Therapies for
Individual ALL Patients

Peter Emanuel, M.D.

We all know that television dramas do not accurately predict what happens in the emergency room nor in crime and murder investigations. But after listening to the Plenary Session presentation of abstract #1, and to Drs. Willman and Downing in the Pediatric ALL Educational Session, maybe the next TV drama should follow the gene expression profiling story. The goal of this detective technique is to improve risk classification schemes in order to precisely tailor therapeutic approaches to individual patients. Such an individual tailoring is felt necessary since 25% of children with ALL will ultimately relapse with resistant disease, yet other subgroups are probably being over-treated with standard therapies and thus at undue risk for long term side effects. Utilizing cooperative group infant leukemia cell/tissue banks, the investigators used unsupervised learning tools as well as supervised machine learning algorithms to identify novel genes predictive of outcome. Amazingly, using Bayesian methods and combining these techniques, a set of only three genes was discovered to be most predictive of outcome in pediatric ALL.

The three genes identified were: (1) G0, a previously uncloned EST; (2) G1, which was GNB2L1, a Gprotein and PKC activator, and (3) G3, the IL-10 receptor að.

G0 was fully cloned by the investigators of plenary abstract #1 and renamed to OPAL1, but not because Dr. Willman and other team investigators liked opal gem stones. OPAL1 stands for Outcome Predictor in Acute Leukemia 1. High OPAL1 gene expression was a remarkably significant predictor for favorable outcome not only for B-cell ALL but also for T-cell ALL, with a similar trend even observed in B precursor ALL. Just listen to some of these statistics: amongst high OPAL1 gene expressors there was an 87% long term remission rate vs. 32% long term remission rate (and 68% treatment failure rate) amongst low OPAL1 gene expressors. And OPAL1 clustered with the good karyotype group of t(12;21), normal karyotype and hyperdiploidy.

Take home message – while women think diamonds are their best friends, children with lymphoid leukemias should dream of lots of opals rather than diamonds. All kidding aside, gene expression profiling and computational modeling are promising, exciting investigative tools that may revolutionize our ability to improve risk classification and outcome prediction in acute leukemia. And after all, that’s what we’re here for. Congratulations to the gene expression profilers.