A molecular assay of tumor radiosensitivity: a roadmap towards biology-based personalized radiation therapy

Abstract

The last two decades have seen technological developments that have led to more accurate delivery of radiation therapy (RT), which has resulted in clinical gains in many solid tumors. However, a fundamental question and perhaps the next major hurdle is whether biological strategies can be developed to further enhance the effectiveness and efficiency of RT. This article addresses the development of a novel genomics-based molecular assay to predict tumor radiosensitivity, and proposes that this assay may prove pivotal in the development of biologically guided RT.

Figure 1. Systems biology approaches to modeling disease

Adapted with permission from [72].

Figure 2. Schema for a radiosensitivity classifier

A leave-one-out cross-validation approach was utilized to develop and test the classifier. Linear regression was utilized to identify radiation-specific biomarkers to build the classifier. The classifier was based on a linear regression equation where y_x stands for expression for the selected genes and k_x are coefficients generated during the training process. Accuracy was 62%, based on continuous variable prediction (not binary; p = 0.002). SF2: Survival fraction at 2 Gy. Data taken from [44]. independent clinical data sets in four disease sites in a total of 621 patients. Importantly, RSI was

Figure 3. Strategy for development of a systems biology model of radiosensitivity

(A) Details the step-by-step approach in the development process. (B) Identification of ten genes in the systems model. A linear regression equation was developed to correlate gene expression and cellular radiosensitivity in a database of 48 cell lines. The final algorithm for RSI was developed in the 48-cell line database as a linear function of gene expression for the ten genes identified. This final lock-down algorithm was then validated in five independent data sets, totaling 621 patients. ^†Data set originated from the Karolinska Institute (Stockholm, Sweden). ^‡Data set originated from the Erasmus Medical Center (Rotterdam, The Netherlands). RSI: Radiosensitivity index; TO: Tissue of origin. Adapted with permission from [45] © Elsevier (2009).

Figure 4. Radiosensitivity index is correlated with clinical response to concurrent radiochemotherapy in rectal and esophageal cancer patients

Radiosensitivity index (RSI) for each patient was generated using a rank-based linear regression model built from the cell line data. Statistical significance was determined using a one-sided Mann–Whitney test for differences. (A) The mean RSI of responders is significantly lower than in nonresponders in both clinical cohorts (esophageal: p = 0.05; rectal: p = 0.03). (B) RSI for each individual patient in the cohorts (combined: p = 0.002). PPV: Positive predictive value; Sens.: Sensitivity; Spec.: Specificity; SF2: Survival fraction at 2 Gy. Reproduced with permission from [64] © Elsevier (2009).

Figure 5. Radiosensitivity index distinguishes clinical populations with different disease-related outcomes in head and neck cancer

Radiosensitivity predictions were generated with the gene-expression model as described in 92 patients treated with definitive concurrent radiochemotherapy at The Netherlands Cancer Institute (Amsterdam, The Netherlands). Using the 25th percentile as the cutoff point, there is a superior 2-year locoregional control in the predicted radiosensitive group (green vs red, 86 vs 61%; p = 0.05) Reproduced with permission from [64] © Elsevier (2009).