Prediction of the outcome of preoperative chemotherapy in breast cancer using DNA probes that provide information on both complete and incomplete responses

Abstract

Background: DNA microarray technology has emerged as a major tool for exploring cancer biology and solving clinical issues. Predicting a patient's response to chemotherapy is one such issue; successful prediction would make it possible to give patients the most appropriate chemotherapy regimen. Patient response can be classified as either a pathologic complete response (PCR) or residual disease (NoPCR), and these strongly correlate with patient outcome. Microarrays can be used as multigenic predictors of patient response, but probe selection remains problematic. In this study, each probe set was considered as an elementary predictor of the response and was ranked on its ability to predict a high number of PCR and NoPCR cases in a ratio similar to that seen in the learning set. We defined a valuation function that assigned high values to probe sets according to how different the expression of the genes was and to how closely the relative proportions of PCR and NoPCR predictions to the proportions observed in the learning set was. Multigenic predictors were designed by selecting probe sets highly ranked in their predictions and tested using several validation sets.

Semi-Supervised Model Applied to the Prediction of the Response to Preoperative Chemotherapy for Breast Cancer

“…The works of Horta (Horta, 2008), Hess (K.R. Hess and Pusztai., 2006), Natowicz (Rene Natowicz and Rouzier., march 2008) and Braga (R. Natowicz and Costa, 2008) (Braga et al, n.d.), have made use of these data to select relevant probes for prediction. Among them, we have selected three probes sets for our application of our semi-supervised learning method : the 30 probes set selected by Natowicz (Rene Natowicz and Rouzier., march 2008), the 18 probes set and the 11 probes set selected by Horta (Horta, 2008) The results ( The best SSL method results are comparable to best results achieved in (R. Natowicz, 2009) that used a SVM with linear kernels (See table 4 of (R. Natowicz, 2009) …”

Bayesian Classifiers for Predicting the Outcome of Breast Cancer Preoperative Chemotherapy

“…In our previous studies ( [4,5]) we had defined the k-probes majority decision predictor as the set of the k top ranked probes together with the majority decision criterion: for any patient case, when the majority of 'pcr' and 'nopcr' predictions of the k top ranked probes was 'pcr', the patient was predicted to be 'PCR', and when the majority was 'nopcr' the patient was predicted to be 'NoPCR'. In case of tie the patient was predicted 'UNSPECIFIED'.…”

“…[5,4]. We assigned two sets of expression levels to any probe s, the sets E p (s) and E n (s), computed from the training data as follows [5]. Let m p (s) et sd p (s) be the mean and standard deviation of the expression levels of probe s for the PCR training cases, and let m n (s) and sd n (s) be that of the NoPCR training cases.…”

“…For each patient case, the data was the outcome of the treatment and the expression levels of 22283 DNA probes of an Affymetrix U133A microarray, measured on tumor cells. [5,4]. We assigned two sets of expression levels to any probe s, the sets E p (s) and E n (s), computed from the training data as follows [5].…”

“…A pathologic complete response (PCR) at surgery is correlated with an positive outcome while residual disease (NoPCR) is associated with a negative outcome. An accurate prediction of tumor sensitivity to preoperative chemotherapy is important to avoid the prescription of an inefficient treatment to patients with predicted residual disease while allocating the treatment to all the PCR patients In [4,5] we have proposed a new method for DNA probes selection. In this paper we investigate the performances of bayesian classifiers taking as input the expression levels of the selected probes.…”