Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2012 Jan 3;106(1):126-32.
doi: 10.1038/bjc.2011.505. Epub 2011 Nov 17.

Potential responders to FOLFOX therapy for colorectal cancer by Random Forests analysis

S Tsuji  1 , Y Midorikawa, T Takahashi, K Yagi, T Takayama, K Yoshida, Y Sugiyama, H Aburatani
Affiliations

Potential responders to FOLFOX therapy for colorectal cancer by Random Forests analysis

S Tsuji et al. Br J Cancer. .

Abstract

Background: Molecular characterisation using gene-expression profiling will undoubtedly improve the prediction of treatment responses, and ultimately, the clinical outcome of cancer patients.

Methods: To establish the procedures to identify responders to FOLFOX therapy, 83 colorectal cancer (CRC) patients including 42 responders and 41 non-responders were divided into training (54 patients) and test (29 patients) sets. Using Random Forests (RF) algorithm in the training set, predictor genes for FOLFOX therapy were identified, which were applied to test samples and sensitivity, specificity, and out-of-bag classification accuracy were calculated.

Results: In the training set, 22 of 27 responders (81.4% sensitivity) and 23 of 27 non-responders (85.1% specificity) were correctly classified. To improve the prediction model, we removed the outliers determined by RF, and the model could correctly classify 21 of 23 responders (91.3%) and 22 of 23 non-responders (95.6%) in the training set, and 80.0% sensitivity and 92.8% specificity, with an accuracy of 69.2% in 29 independent test samples.

Conclusion: Random Forests on gene-expression data for CRC patients was effectively able to stratify responders to FOLFOX therapy with high accuracy, and use of pharmacogenomics in anticancer therapy is the first step in planning personalised therapy.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Flow diagram for the present study. Random Forests analysis was totally conducted four times. The final model with 15 probe sets was used for predicting the response of the independent 29 samples.
Figure 2
Figure 2
Classification accuracy of responders to FOLFOX therapy. (A and B) Classification accuracy of responders to FOLFOX therapy using 50 top-ranked genes selected by Random Forests in the training set. (A) Probabilities of sensitivity for FOLFOX therapy in out-of-bag cross-validation. Cutoff value was defined as response rate, 0.5. In all, 22 of 27 sensitive patients (81.4% sensitivity) and 23 of 27 resistant patients (85.1% specificity) were correctly classified, with an accuracy of 62.1% (blue square, responder; red triangle, non-responder). (B) Proximity matrix by predictor genes for FOLFOX therapy. At the intersection of each column and row in the figure is a pixel, the intensity of which is a measure of the distance (defined as 1−Peason's correlation coefficient) between the centroids named by the intersecting column and row. The red area corresponds to a high degree of co-occurrence, that is, these samples tend to cluster in all clustering runs. Asterisk on the patient numbers indicate outliers. Red, blue, and black boxes below the proximity matrix represent non-responder, responder, and outlier, respectively. (C and D) Classification accuracy of responders for FOLFOX therapy using 15 predictor probes (14 genes) after removing 8 outliers from the training set. (C) Probabilities of sensitivity for FOLFOX therapy in out-of-bag cross-validation after removing 8 outliers. Sensitivity (91.3%), specificity (95.6%), and out-of-bag classification accuracy (80.2%) were markedly improved. (D) Proximity matrix by predictor genes for FOLFOX therapy after removing 8 outliers. Outlier scores were calculated again in 46 samples, all of which were <6.0.
Figure 3
Figure 3
Predicted probabilities using 14 predictor genes for FOLFOX therapy in test samples. Using the prediction model in the training set, 12 of 15 sensitive patients (80.0% sensitivity) and 13 of 14 resistant patients (92.8% specificity) were correctly classified, with an averaged accuracy rate of 69.2% in the test set. The order of samples in A correspond to the B. (A) The heat map of the expression values of 14 predictor genes. As NPEPPS has two probes, the heat map has 15 rows. (B) Predicted response probability of 29 test samples (blue square, responder; red triangle, non-responder).
Figure 4
Figure 4
Overall survival of unresectable colorectal cancer patients. The response signature was used to predict overall survival in a training set (A) and a test set (B) of unresectable CRC patients treated with FOLFOX therapy. The predicted probability of the signature was used to identify individual patients exhibiting the phenotype. Continuous line, patients determined as responders by Random Forests algorithm; broken line, non-responders.
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Affymetrix. Technical notes. Affymetrix I (2005) Guide to probe logarithmic intensity error (plier) estimation. http://www.affymetrix.com/support/technical/technotesmain.affx
    1. Aster JC, Xu L, Karnell FG, Patriub V, Pui JC, Pear WS (2000) Essential roles for ankyrin repeat and transactivation domains in induction of T-cell leukemia by notch1. Mol Cell Biol 20: 7505–7515 - PMC - PubMed
    1. Berchuck A, Iversen ES, Lancaster JM, Pittman J, Luo J, Lee P, Murphy S, Dressman HK, Febbo PG, West M, Nevins JR, Marks JR (2005) Patterns of gene expression that characterize long-term survival in advanced stage serous ovarian cancers. Clin Cancer Res 11: 3686–3696 - PubMed
    1. Braun MS, Adab F, Bradley C, McAdam K, Thomas G, Wadd NJ, Rea D, Philips R, Twelves C, Bozzino J, MacMillan C, Saunders MP, Counsell R, Anderson H, McDonald A, Stewart J, Robinson A, Davies S, Richards FJ, Seymour MT (2003) Modified de Gramont with oxaliplatin in the first-line treatment of advanced colorectal cancer. Br J Cancer 89: 1155–1158 - PMC - PubMed
    1. Breiman L (2001) Random Forests. Mach Learn 45: 5–32

Publication types

MeSH terms

Supplementary concepts