Personalized chemotherapy selection for breast cancer using gene expression profiles

Abstract

Choosing the optimal chemotherapy regimen is still an unmet medical need for breast cancer patients. In this study, we reanalyzed data from seven independent data sets with totally 1079 breast cancer patients. The patients were treated with three different types of commonly used neoadjuvant chemotherapies: anthracycline alone, anthracycline plus paclitaxel, and anthracycline plus docetaxel. We developed random forest models with variable selection using both genetic and clinical variables to predict the response of a patient using pCR (pathological complete response) as the measure of response. The models were then used to reassign an optimal regimen to each patient to maximize the chance of pCR. An independent validation was performed where each independent study was left out during model building and later used for validation. The expected pCR rates of our method are significantly higher than the rates of the best treatments for all the seven independent studies. A validation study on 21 breast cancer cell lines showed that our prediction agrees with their drug-sensitivity profiles. In conclusion, the new strategy, called PRES (Personalized REgimen Selection), may significantly increase response rates for breast cancer patients, especially those with HER2 and ER negative tumors, who will receive one of the widely-accepted chemotherapy regimens.

Genome-wide analysis of therapeutic response uncovers molecular pathways governing tamoxifen resistance in ER+ breast cancer

“…[ 30 ] method, which used Support Vector Machine approach as a base; (iii) Yu et al. [ 31 ] method, also referred to as Personalized REgimen Selection (PRES), which used random forest approach as a base; and (iv) mRNA expression alone (without taking into account information about molecular pathways) ( see Materials and Methods ). To assure that all methods are comparable to our pathway-centric method, we trained Epsi et al., Zhong et al., Yu et al., and expression-only methods on the Training cohort, with each producing a list of predictions, either pathways or gene lists, depending on the method (112 predictions for Epsi et al.…”

“…To assess the advantages of our approach over other commonly used techniques, we compared its performance to (i) extreme-responder analysis, described in Epsi et al. [ 28 ]; (ii) SVM-based method [ 30 ]; (iii) PRES random forest-based method [ 31 ]; and (iv) expression data alone (without utilizing biological pathways). In each case, we utilized Training cohort for model training and Test cohort 1 for model validation.…”

“…[ 30 ] (which used Support Vector Machine approach as a base), Yu et al. [ 31 ] (which uses random forest approach as a base), and mRNA data alone (without considering molecular pathways) and demonstrated that our method outperforms these techniques in predicting risk of resistance to tamoxifen. Furthermore, we have compared our pathway signature to other known signatures of tamoxifen response [18] , [19] , [20] and have shown the superiority of our pathway-based approach (adjusted hazard ratio = 3.11, hazard p-value = 0.0278).…”

Correlation-centred variable selection of a gene expression signature to predict breast cancer metastasis

“…As statistical approaches, multivariable logistic regression has been used for binary outcomes such as metastasis 16 . As machine-learning methods, artificial neural networks, Bayesian networks, support vector machines (SVMs), and decision trees (DTs) are typical methods for binary outcomes 17 , and machine learning-based analyses have been applied to the prediction of cancer prognosis [17][18][19] . Combining multiple methods has also been reported, such as ReliefF-GA-ANFIS 18 .…”

Maximizing the Reusability of Public Gene Expression Data by Predicting Missing Metadata

“…It has been shown that gene expression is very informative and highly predictive of various clinical outcomes, such as the status of biomarkers [15][16][17][18][19][20][21], tumor types/status [22][23][24][25], the risks of recurrence [26,27] and survival [27][28][29][30], and therapeutic response [4,[31][32][33][34]. For gene expression data, the gene expression profiles themselves, therefore, are ideally suited for inferring the missing metadata.…”