Keywords
Gene expression signatures, breast cancer, chemotherapy resistance, hormone therapy, machine learning, support vector machine, random forest
This article is included in the Bioinformatics gateway.
This article is included in the Machine learning: life sciences collection.
Gene expression signatures, breast cancer, chemotherapy resistance, hormone therapy, machine learning, support vector machine, random forest
Changes to the manuscript have been incorporated in response to the valuable comments provided by Drs. Fertig and Tung. These include three additional tables, which report the results for cross-validation using the METABRIC Discovery set only. We have included Supplementary Material demonstrating heterogeneous expression of paclitaxel signature genes in the Discovery vs. Validation datasets. Methods have been updated with additional detail as requested. Results now include additional performance metrics. We have also corrected some predictions in Dataset 1, however changes do not affect the conclusions of the paper.
See the authors' detailed response to the review by Elana J. Fertig
See the authors' detailed response to the review by Chun-Wei Tung
Current pharmacogenetic analysis of chemotherapy makes qualitative decisions about drug efficacy in patients (determination of good, intermediate or poor metabolizer phenotypes) based on variants present in genes involved in the transport, biotransformation, or disposition of a drug. We have applied a supervised machine learning (ML) approach to derive accurate gene signatures, based on the biochemically-guided response to chemotherapies with breast cancer cell lines1, which show variable responses to growth inhibition by paclitaxel and gemcitabine therapies2,3. We analyzed stable4 and linked unstable genes in pathways that determine their disposition. This involved investigating the correspondence between 50% growth inhibitory concentrations (GI50) of paclitaxel and gemcitabine and gene copy number, mutation, and expression first in breast cancer cell lines and then in patients1. Genes encoding direct targets of these drugs, metabolizing enzymes, transporters, and those previously associated with chemo-resistance to paclitaxel (n=31 genes) were then pruned by multiple factor analysis (MFA), which indicated that expression levels of genes ABCC10, BCL2, BCL2L1, BIRC5, BMF, FGF2, FN1, MAP4, MAPT, NKFB2, SLCO1B3, TLR6, TMEM243, TWIST1, and CSAG2 could predict sensitivity in breast cancer cell lines with 84% accuracy. The cell line-based paclitaxel-gene signature predicted sensitivity in 84% of patients with no or minimal residual disease (n=56; data from 5). The present study derives related gene signatures with ML approaches that predict outcome of hormone- and chemotherapies in the large METABRIC breast cancer cohort6.
SVM learning: Previously, paclitaxel-related response genes were identified from peer-reviewed literature, and their expression and copy number in breast cancer cell lines were analyzed by multiple factor analysis of GI50 values of these lines2 (Figure 1). Given the expression levels of each gene, a SVM is evaluated on patients by classifying those with shorter survival time as resistant and longer survival as sensitive to hormone and/or chemotherapy using paclitaxel, tamoxifen, methotrexate, 5-fluorouracil, epirubicin, and doxorubicin. The SVM was trained using the function fitcsvm in MATLAB R2014a7 and tested with either leave-one-out or 9 fold cross-validation (indicated in Table 1). The Gaussian kernel was used for this study, unlike Dorman et al.1 which used the linear kernel. The SVM requires selection of two different parameters, C (misclassification cost) and sigma (which controls the flexibility and smoothness of Gaussians)8; these parameters determine how strictly the SVM learns the training set, and hence if not selected properly, can lead to overfitting. A grid search evaluates a wide range of combinations of these values by parallelization. A Gaussian kernel selects the C and sigma combination that lead to the lowest cross-validation misclassification rate. A backwards feature selection (greedy) algorithm was designed and implemented in MATLAB in which one gene of the set is left out in a reduced gene set and the classification is then assessed; genes that maintain or lower the misclassification rate are kept in the signature. The procedure is repeated until the subset with the lowest misclassification rate is selected as the optimal subset of genes. These SVMs were then assessed for their ability to predict patient outcomes based on available metadata (see Figure 1 and reference 1). Interactive prediction using normalized expression values as input is available at http://chemotherapy.cytognomix.com.
Patient treatment | # of patients | Agent: final gene signature (C and sigma) | Accuracy (%) | Precision | F-Measure | MCC1 | AUC2 |
---|---|---|---|---|---|---|---|
Both CT and HT3 | 84 | Paclitaxel: ABCC1, ABCC10, BAD, BIRC5, FN1, GBP1, MAPT, SLCO1B3, TMEM243, TUBB3, TUBB4B (C=10000, σ=10) | 78.6 | 0.787 | 0.782 | 0.559 | 0.814 |
Tamoxifen: ABCC2, ALB, CCNA2, E2F7, FLAD1, FMO1, NCOA2, NR1I2, PIAS4, SULT1E1 (C=100000, σ=100) | 76.2 | 0.761 | 0.760 | 0.510 | 0.701 | ||
Methotrexate: ABCC2, ABCG2, CDK2, DHFRL1 (C=10, σ=1) | 71.4 | 0.712 | 0.711 | 0.410 | 0.766 | ||
Epirubicin: ABCB1, CDA, CYP1B1, ERBB3, ERCC1, MTHFR, PON1, SEMA4D, TFDP2 (C=1000, σ=10) | 72.6 | 0.725 | 0.723 | 0.434 | 0.686 | ||
Doxorubicin: ABCC2, ABCD3, CBR1, FTH1, GPX1, NCF4, RAC2, TXNRD1 (C=100000, σ=100) | 75.0 | 0.749 | 0.750 | 0.488 | 0.701 | ||
5-Fluorouracil: ABCB1, ABCC3, MTHFR, TP53 (C=10000, σ=100) | 71.4 | 0.714 | 0.714 | 0.417 | 0.718 | ||
CT and/or HT3,4,5,6 | 735 | Paclitaxel: BAD, BCAP29, BCL2, BMF, CNGA3, CYP2C8, CYP3A4, FGF2, FN1, NFKB2, NR1I2, OPRK1, SLCO1B3, TLR6, TUBB1, TUBB3, TUBB4A, TUBB4B, TWIST1 (C=10000, σ=100) | 66.1 | 0.652 | 0.643 | 0.287 | 0.660 |
Deceased only2,6,7 (CT and/or HT) | 327 | Paclitaxel: ABCB11, BAD, BBC3, BCL2, BCL2L1, BIRC5, CYP2C8, FGF2, FN1, GBP1, MAPT, NFKB2, OPRK1, SLCO1B3, TMEM243 (C=100, σ=10) | 75.3 | 0.752 | 0.752 | 0.505 | 0.763 |
No treatment3 | 304 | Paclitaxel: ABCB1, ABCB11, BBC3, BCL2L1, BMF, CYP3A4, FGF2, GBP1, MAP4, MAPT, NR1I2, OPRK1, SLCO1B3, TUBB4A, TUBB4B, TWIST2 (C=100, σ=10) | 73.4 | 0.734 | 0.733 | 0.467 | 0.769 |
Initial gene sets preceding feature selection: Paclitaxel - ABCB1, ABCB11, ABCC1, ABCC10, BAD, BBC3, BCAP29, BCL2, BCL2L1, BIRC5, BMF, CNGA3, CYP2C8, CYP3A4, FGF2, FN1, GBP1, MAP2, MAP4, MAPT, NFKB2, NR1I2, OPRK1, SLCO1B3, TLR6, TUBB1, TWIST1. Tamoxifen - ABCB1, ABCC2, ALB, C10ORF11, CCNA2, CYP3A4, E2F7, F5, FLAD1, FMO1, IGF1, IGFBP3, IRS2, NCOA2, NR1H4, NR1I2, PIAS4, PPARA, PROC, RXRA, SMARCD3, SULT1B1, SULT1E1, SULT2A1. Methotrexate - ABCB1, ABCC2, ABCG2, CDK18, CDK2, CDK6, CDK8, CENPA, DHFRL1. Epirubicin - ABCB1, CDA, CYP1B1, ERBB3, ERCC1, GSTP1, MTHFR, NOS3, ODC1, PON1, RAD50, SEMA4D, TFDP2. Doxorubicin - ABCB1, ABCC2, ABCD3, AKR1B1, AKR1C1, CBR1, CYBA, FTH1, FTL, GPX1, MT2A, NCF4, RAC2, SLC22A16, TXNRD1. 5-Fluorouracil - ABCB1, ABCC3, CFLAR, IL6, MTHFR, TP53, UCK2. 1MCC: Matthews Correlation Coefficient. 2AUC: Area under receiver operating curve. 3 Surviving patients; 4 Analysis included patients in the METABRIC ‘discovery’ dataset only; 5 SVMs tested with 9 fold cross-validation, all others tested with leave-one-out cross-validation; 6 Includes all patients treated with HT,CT, combination CT/HT, either with or without combination radiotherapy; 7 Median time after treatment until death (> 4.4 years) was used to distinguish favorable outcome, ie. sensitivity to therapy.
RF learning: RF was trained using the WEKA 3.79 data mining tool. This classifier uses multiple random trees for classification, which are combined via a voting scheme to make a decision on the given input gene set. A grid search was used to optimize the maximum number of randomly selected genes for each tree in RF, where k (maximum number of selected genes for each tree) was set from 1 to 19. Figure 2 depicts the therapy outcome prediction process of a given patient using a RF consisting of a series of decision trees derived from different subsets of paclitaxel-related genes.
Augmented Gene Selection: The most relevant genes (features) for therapy outcome prediction were found using the Minimum Redundancy and Maximum Relevance (mRMR) approach10. mRMR is a wrapper approach that incrementally selects genes by maximizing the average mutual information between gene expression features and classes, while minimizing their redundancies:
where fi corresponds to a feature in gene set S, I(fi,C) is the mutual information between fi and class C, and I(fi,fj) is the mutual information between features fi and fj.
For this experiment, we used a 26-gene signature (genes ABCB1, ABCB11, ABCC1, ABCC10, BAD, BBC3, BCL2, BCL2L1, BMF, CYP2C8, CYP3A4, MAP2, MAP4, MAPT, NR1I2, SLCO1B3, TUBB1, TUBB4A, TUBB4B, FGF2, FN1, GBP1, NFKB2, OPRK1, TLR6, and TWIST1) as the base feature set. These genes were selected (in Dorman et al.1) based either on their known involvement in paclitaxel metabolism, or evidence that their expression levels and/or copy numbers correlate with paclitaxel GI50 values. mRMR and SVM were combined to obtain a subset of genes that can accurately predict patient survival outcomes; here, we considered 3, 4 and 5 years as survival thresholds for breast cancer patients.
Performance was evaluated with several metrics. WEKA determined accuracy (ACC), the weighted average of precision and F-measure, the Matthews Correlation Coefficient (MCC) and the area under ROC curve (AUC).
Type of treatment | Survival years (as threshold) | # Patients | K (number of genes to be used in random selection) | Accuracy (True Positive - TP) (%) | Precision | F-Measure | MCC1 | AUC2 |
---|---|---|---|---|---|---|---|---|
Chemotherapy (CT) | 3 | 53 | 7 | 56.6 | 0.510 | 0.524 | -0.059 | 0.441 |
4 | 7 | 69.8 | 0.698 | 0.698 | 0.396 | 0.700 | ||
5 | 19 | 66.0 | 0.645 | 0.636 | 0.230 | 0.653 | ||
Hormone therapy (HT) | 3 | 420 | 19 | 85.5 | 0.731 | 0.788 | 0.000 | 0.606 |
4 | 9 | 78.6 | 0.715 | 0.706 | 0.069 | 0.559 | ||
5 | 9 | 71.0 | 0.634 | 0.627 | 0.059 | 0.632 | ||
CT and/or HT | 3 | 504 | 9 | 82.7 | 0.685 | 0.749 | 0.000 | 0.506 |
4 | 19 | 73.6 | 0.647 | 0.648 | 0.039 | 0.527 | ||
5 | 7 | 65.3 | 0.602 | 0.593 | 0.086 | 0.588 |
1MCC: Matthews Correlation Coefficient. 2AUC: Area under receiver operating curve; both Discovery and Validation patient datasets analyzed. RF predictions done using a gene panel consisting of 19 genes (ABCB1, ABCB11, ABCC1, ABCC10, BAD, BBC3, BCL2, BCL2L1, BMF, CYP2C8, CYP3A4, MAP2, MAP4, MAPT, NR1I2, SLCO1B3, TUBB1, TUBB4A, TUBB4B).
Data | CT1 | HT | CT+HT | ||||||
---|---|---|---|---|---|---|---|---|---|
Survival years (as threshold) | 3 | 4 | 5 | 3 | 4 | 5 | 3 | 4 | 5 |
# patients2 | 53 | 420 | 504 | ||||||
Accuracy (TP) (%) | 81.1 | 81.1 | 84.9 | 85.7 | 79.5 | 72.9 | 83.1 | 74.8 | 67.9 |
Precision | 0.809 | 0.813 | 0.852 | 0.878 | 0.765 | 0.692 | 0.795 | 0.703 | 0.662 |
F-Measure | 0.809 | 0.811 | 0.845 | 0.794 | 0.726 | 0.663 | 0.772 | 0.672 | 0.666 |
MCC | 0.582 | 0.625 | 0.675 | 0.119 | 0.17 | 0.173 | 0.161 | 0.137 | 0.238 |
AUC | 0.783 | 0.812 | 0.82 | 0.508 | 0.533 | 0.548 | 0.53 | 0.531 | 0.61 |
SVM Par. (gamma) | 0.0 | 0.5 | 1.0 | 1.0 | 0.75 | 1.5 | 0.75 | 0.5 | 1.0 |
SVM Par. (cost) | 64 | 128 | 8 | 2 | 64 | 2 | 16 | 2 | 2 |
Selected genes | MAP4, GBP1, FN1, MAPT, BBC3, FGF2, NFKB2, TUBB4B | TWIST1, FN1, BBC3, FGF2, BCL2L1 | ABCB11, BCL2, GBP1, SLCO1B3, ABCB1, BAD, TUBB4A, MAPT, NFKB2, TUBB4B | ABCB11, BCL2, MAP4, TUBB1, GBP1, SLCO1B3, ABCB1, BAD, TWIST1, FN1, TUBB4A, MAPT, OPRK1, BBC3, FGF2, NFKB2, ABCC1, NR1I2 | BAD, GBP1, MAPT, BBC3 | ABCB11, MAP4, SLCO1B3, BAD, FN1, OPRK1, BBC3, NFKB2, NR1I2, TUBB4B | ABCB11, SLCO1B3, BAD, TUBB4A, MAPT, BBC3, FGF2, NFKB2, ABCC1, NR1I2 | ABCB11, BMF, BCL2, MAP4, TUBB1, GBP1, SLCO1B3, ABCB1, BAD, TWIST1, FN1, MAPT, OPRK1, BBC3, FGF2, NFKB2, ABCC1, NR1I2, TUBB4B | MAP4, GBP1, SLCO1B3, BAD, MAPT, OPRK1, BBC3, NFKB2, ABCC1, NR1I2, TUBB4B |
1 For patients treated with CT with ≥4 Yr survival and CT+ HT for ≥ 5 Yr, the cost for the mRMR model was set to 64. Of those treated with CT for ≥ 4 Yr, genes were selected using a greedy, stepwise forward search, while in other cases, greedy stepwise backward search was used. Also, gamma = 0 in all cases. 2Predicted responses for individual METABRIC patients are provided in Dataset 1.
The performances of several ML techniques have been compared such that they distinguish paclitaxel sensitivity and resistance in METABRIC patients using its tumour gene expression datasets. We used mRMR to generate gene signatures and determine which genes are important for treatment response in METABRIC patients. The paclitaxel models are more accurate for prediction of outcomes in patients receiving HT and/or CT compared to other patient groups.
SVMs and RF were trained using expression of genes associated with paclitaxel response, mechanism of action and stable genes in the biological pathways of these targets (Figure 3). Pair-wise comparisons of these genes with those from MammaPrint and Oncotype Dx (other genomic classifiers for breast cancer) find that these signatures are nearly independent of each other, with only a single gene overlap. The distinct differences of these signatures are due to their methodology of derivation, based on different principles and for different purposes (i.e. drug response for a specific reagent). SVM models for drugs used to treat these patients were derived by backwards feature selection on patient subsets stratified by treatment or outcome (Table 1). The highest SVM accuracy was found for the paclitaxel signature in patients treated with HT and/or adjuvant chemotherapy (78.6%). Since some CT patients were also treated with tamoxifen, methotraxate, epirubicin, doxorubicin and 5-fluorouracil, we also evaluated the performance of models developed for these drugs using the same algorithm. These gene signatures also had acceptable performance (accuracies between 71–76%; AUCs between 0.686 – 0.766). Leave-one-out validation (CT and HT, no treatment, and deceased patients) exhibited higher model performance than 9-fold crossvalidation (CT and/or HT, including patients treated with radiation).
The RF classifier was used to predict paclitaxel therapy outcome for patients that underwent CT and/or HT (Table 2). The best performance achieved with RF showed an 85.5% overall accuracy using a 3-year survival threshold for distinguishing therapeutic resistance vs. sensitivity for those patients that underwent HT.
The best overall accuracy and AUC (sensitivity and specificity) for CT/HT patients using mRMR feature selection for SVM predicting outcome of paclitaxel therapy was obtained for CT patients with 4-year survival (Table 3). Outcomes for HT patients with 3-year survival were predicted with 85.7% accuracy; however, the specificity was lower in this group. SVM combined with mRMR further improved accuracy of feature selection and prediction of response to hormone and/or chemotherapy based on survival time than either SVM or RF alone. Predicted treatment responses for individual METABRIC patients using the described ML techniques are indicated in Dataset 1.
We also assessed the separate Discovery and Validation datasets as training and test sets, respectively, and repeated the previous experiments. In this scenario, the performance of the model was poor (slightly better than random). This occurred because the gene expression distributions of many of the paclitaxel-related genes in our signature were not reproducible between these two sets and were in fact, quite different (based on Wilcoxon rank sum test, Kruskal-Wallis test and t-tests; Supplementary file 1). This heterogeneity indicates that it is inappropriate to test our gene expression signatures derived by one of these datasets using the other dataset. Furthermore, these gene expression differences also affect the performance of these methods (compare Table 2 and Table 4 for RF; Table 3 and Table 5 for mRMR).
To evaluate the paclitaxel models without relying on the Validation dataset, the Discovery set was split into two distinct parts, consisting of 70% of the patient samples randomly selected for training, and a different set of 30% of samples for testing. This procedure was repeated 100 times using different combinations of training and test samples, and the median performance of these runs is reported (Table 4 and Table 5). We also compared the performance of our mRMR+SVM model with the K-TSP model (presented in 11; Table 6). In most cases, our method outperformed K-TSP, based on its accuracy in classifying new patients.
Treatment | CT1 | HT | CT+HT | ||||||
---|---|---|---|---|---|---|---|---|---|
Survival years (as threshold) | 3 | 4 | 5 | 3 | 4 | 5 | 3 | 4 | 5 |
# patients | 22 | 185 | 221 | ||||||
Accuracy (TP) (%) | 57.14 | 57.14 | 85.7 | 81.8 | 70.9 | 63.6 | 71.2 | 69.7 | 71.2 |
Precision | 0.595 | 0.686 | 0.735 | 0.726 | 0.670 | 0.532 | 0.647 | 0.629 | 0.693 |
F-Measure | 0.571 | 0.623 | 0.791 | 0.769 | 0.686 | 0.562 | 0.668 | 0.628 | 0.666 |
MCC | 0.167 | -0.258 | 0.000 | -0.080 | 0.032 | -0.075 | 0.035 | 0.071 | 0.245 |
AUC | 0.583 | 0.333 | 0.500 | 0.479 | 0.514 | 0.477 | 0.513 | 0.521 | 0.586 |
SVM Par. (gamma) | 0.0 | 0.5 | 1.0 | 1.0 | 0.75 | 1.5 | 0.75 | 0.5 | 1.0 |
SVM Par. (cost) | 64 | 128 | 8 | 2 | 64 | 2 | 16 | 2 | 2 |
Selected genes | TWIST1 BMF CYP2C8 CYP3A4 BCL2L1 BBC3 BAD MAP2 MAPT NFKB2 FN1 | BCL2 BMF CYP2C8 CYP3A4 BAD ABCC10 NFKB2 | MAP2 BCL2 BCL2L1 BBC3 MAPT GBP1 NFKB2 | TWIST1 BCL2 BMF CYP2C8 CYP3A4 BCL2L1 BBC3 TLR6 BAD ABCB11 ABCC1 ABCC10 MAP4 MAPT NR1I2 GBP1 NFKB2 OPRK1 FN1 | TWIST1 CYP2C8 CYP3A4 BCL2L1 BBC3 TLR6 ABCB11 ABCC1 ABCC10 MAP2 MAPT NR1I2 GBP1 NFKB2 FN1 | TWIST1 BMF CYP2C8 CYP3A4 BCL2L1 BBC3 ABCB11 ABCC1 ABCC10 MAP2 MAP4 MAPT NR1I2 GBP1 NFKB2 OPRK1 | BMF CYP2C8 BCL2L1 BBC3 BAD ABCC1 ABCC10 MAP4 NR1I2 GBP1 NFKB2 OPRK1 FN1 | TWIST1 BMF CYP2C8 CYP3A4 BCL2L1 BBC3 TLR6 ABCB11 ABCC1 ABCC10 MAP2 MAP4 MAPT NR1I2 GBP1 NFKB2 OPRK1 FN1 | TWIST1 BMF CYP3A4 BCL2L1 BBC3 TLR6 BAD ABCB11 ABCC1 MAP2 MAP4 MAPT NR1I2 GBP1 NFKB2 OPRK1 FN1 |
1For patients treated with CT with ≥4 Yr survival and CT+ HT for ≥ 5 Yr, the cost for the mRMR model was set to 64. Of those treated with CT for ≥ 4 Yr, genes were selected using a greedy, stepwise forward search, while in other cases, greedy stepwise backward search was used. Also, gamma = 0 in all cases.
Data | CT | HT | CT+HT | ||||||
---|---|---|---|---|---|---|---|---|---|
Survival years | 3 | 4 | 5 | 3 | 4 | 5 | 3 | 4 | 5 |
# patients | 22 | 185 | 221 | ||||||
mRMR+SVM Accuracy (%) | 57.14 | 57.14 | 85.7 | 81.8 | 70.9 | 63.6 | 71.21 | 69.70 | 71.21 |
K-TSP11 Accuracy (%) | 57.14 | 28.57 | 28.57 | 80.91 | 68.18 | 69.19 | 71.21 | 54.55 | 53.03 |
The performances of several ML techniques have been compared such that they distinguish paclitaxel sensitivity and resistance in METABRIC patients using its tumour gene expression datasets. We used mRMR to generate gene signatures and determine which genes are important for treatment response in METABRIC patients. The paclitaxel models are more accurate for prediction of outcomes in patients receiving HT and/or CT compared to other patient groups.
While not a replication study sensu stricto, the initial paclitaxel gene set used for feature selection was the same as in our previous study1. Predictions for the METABRIC patient cohort, which was independent of the previous validation set5 used in Dorman et al.1, of the either same (SVM) or different ML methods (RF and SVM with mRMR) exhibited comparable or better accuracies than our previous gene signature1.
These techniques are powerful tools which can be used to identify genes that may be involved in drug resistance, as well as predict patient survival after treatment. Future efforts to expand these models to other drugs may assist in suggesting preferred treatments in specific patients, with the potential impact of improving efficacy and reducing duration of therapy.
In this study we used METABRIC dataset to predict outcome for different survival times in patients receiving hormone (HT) and, in some cases, chemotherapy (CT) agents. We used different machine learning methods in order to identify the best subset of genes that can accurately predict therapeutic response in patients undergone chemotherapy, hormone therapy or a combination of both treatments. Unlike Mammaprint and Oncotype Dx tests, this model focuses on predicting survival prediction based on gene expression in the tumor, presumably before or during drug therapy. This approach may be useful for selecting specific therapies in patients that would be expected to produce a favorable response.
Patient data: The METABRIC datasets are accessible from the European Genome-Phenome Archive (EGA) using the accession number EGAS00000000083 (https://www.ebi.ac.uk/ega/studies/EGAS00000000083). Normalized patient expression data for the Discovery (EGAD00010000210) and Validation sets (EGAD00010000211) were retrieved with permission from EGA. Corresponding clinical data was obtained from the literature6. While not individually curated, HT patients were treated with tamoxifen and/or aromatase inhibitors, while CT patients were most commonly treated with cyclophosphamide-methotrexate-fluorouracil (CMF), epirubicin-CMF, or doxorubicin-cyclophosphamide.
F1000Research: Dataset 1. Predicted treatment response for each individual METABRIC patient, 10.5256/f1000research.9417.d14986412
PKR, AN and LR designed the methodology and oversaw the project. SVM feature selection with MATLAB was automated by DA. EJM and KB selected the initial gene signatures, and performed processing of the METABRIC data using SVM methods. IR performed the preprocessing of the METABRIC dataset using RF; IR and HQ designed feature selection and classification modules using WEKA. PKR, IR, EJM, AN, and LR wrote the manuscript.
PKR cofounded Cytognomix. A patent application related to biologically inspired gene signatures is pending. The other authors declare that they have no competing interests.
AN and LR are funded by NSERC grants RGPIN-2016-05017 and RGPIN-2014-05084 and by the Windsor Essex County Cancer Centre Foundation under a Seeds4Hope grant. PKR has been supported by NSERC [Discovery Grant RGPIN-2015-06290], Canadian Foundation for Innovation, Canada Research Chairs and Cytognomix Inc.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Variation of Gene Expression Distribution between Discovery and Validation Datasets.
Whisker plots showing the distribution of expression in the Discovery and Validation METABRIC datasets for 26 genes used in the paclitaxel gene signature.
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Competing Interests: No competing interests were disclosed.
References
1. Curtis C, Shah SP, Chin SF, Turashvili G, et al.: The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups.Nature. 2012; 486 (7403): 346-52 PubMed Abstract | Publisher Full TextCompeting Interests: No competing interests were disclosed.
Competing Interests: No competing interests were disclosed.
Competing Interests: No competing interests were disclosed.
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | ||
---|---|---|
1 | 2 | |
Version 3 (revision) 12 May 17 |
read | |
Version 2 (revision) 27 Jan 17 |
read | read |
Version 1 31 Aug 16 |
read | read |
Click here to access the data.
Spreadsheet data files may not format correctly if your computer is using different default delimiters (symbols used to separate values into separate cells) - a spreadsheet created in one region is sometimes misinterpreted by computers in other regions. You can change the regional settings on your computer so that the spreadsheet can be interpreted correctly.
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)