Dataset Information

Pareto Optimization Identifies Diverse Set of Phosphorylation Signatures Predicting Response to Treatment with Dasatinib.

ABSTRACT: Multivariate biomarkers that can predict the effectiveness of targeted therapy in individual patients are highly desired. Previous biomarker discovery studies have largely focused on the identification of single biomarker signatures, aimed at maximizing prediction accuracy. Here, we present a different approach that identifies multiple biomarkers by simultaneously optimizing their predictive power, number of features, and proximity to the drug target in a protein-protein interaction network. To this end, we incorporated NSGA-II, a fast and elitist multi-objective optimization algorithm that is based on the principle of Pareto optimality, into the biomarker discovery workflow. The method was applied to quantitative phosphoproteome data of 19 non-small cell lung cancer (NSCLC) cell lines from a previous biomarker study. The algorithm successfully identified a total of 77 candidate biomarker signatures predicting response to treatment with dasatinib. Through filtering and similarity clustering, this set was trimmed to four final biomarker signatures, which then were validated on an independent set of breast cancer cell lines. All four candidates reached the same good prediction accuracy (83%) as the originally published biomarker. Although the newly discovered signatures were diverse in their composition and in their size, the central protein of the originally published signature - integrin ?4 (ITGB4) - was also present in all four Pareto signatures, confirming its pivotal role in predicting dasatinib response in NSCLC cell lines. In summary, the method presented here allows for a robust and simultaneous identification of multiple multivariate biomarkers that are optimized for prediction performance, size, and relevance.

SUBMITTER: Klammer M

PROVIDER: S-EPMC4470654 | biostudies-literature | 2015

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Pareto Optimization Identifies Diverse Set of Phosphorylation Signatures Predicting Response to Treatment with Dasatinib.

Klammer Martin M Dybowski J Nikolaj JN Hoffmann Daniel D Schaab Christoph C

PloS one 20150617 6

Multivariate biomarkers that can predict the effectiveness of targeted therapy in individual patients are highly desired. Previous biomarker discovery studies have largely focused on the identification of single biomarker signatures, aimed at maximizing prediction accuracy. Here, we present a different approach that identifies multiple biomarkers by simultaneously optimizing their predictive power, number of features, and proximity to the drug target in a protein-protein interaction network. To ...[more]

PMID: 26083411

Similar Datasets

Project description:A major challenge in drug development is safety and toxicity concerns due to drug side effects. One such side effect, drug-induced liver injury (DILI), is considered a primary factor in regulatory clearance. The Critical Assessment of Massive Data Analysis (CAMDA) 2020 CMap Drug Safety Challenge goal was to develop prediction models based on gene perturbation of six preselected cell-lines (CMap L1000), extended structural information (MOLD2), toxicity data (TOX21), and FDA reporting of adverse events (FAERS). Four types of DILI classes were targeted, including two clinically relevant scores and two control classifications, designed by the CAMDA organizers. The L1000 gene expression data had variable drug coverage across cell lines with only 247 out of 617 drugs in the study measured in all six cell types. We addressed this coverage issue by using Kru-Bor ranked merging to generate a singular drug expression signature across all six cell lines. These merged signatures were then narrowed down to the top and bottom 100, 250, 500, or 1,000 genes most perturbed by drug treatment. These signatures were subject to feature selection using Fisher's exact test to identify genes predictive of DILI status. Models based solely on expression signatures had varying results for clinical DILI subtypes with an accuracy ranging from 0.49 to 0.67 and Matthews Correlation Coefficient (MCC) values ranging from -0.03 to 0.1. Models built using FAERS, MOLD2, and TOX21 also had similar results in predicting clinical DILI scores with accuracy ranging from 0.56 to 0.67 with MCC scores ranging from 0.12 to 0.36. To incorporate these various data types with expression-based models, we utilized soft, hard, and weighted ensemble voting methods using the top three performing models for each DILI classification. These voting models achieved a balanced accuracy up to 0.54 and 0.60 for the clinically relevant DILI subtypes. Overall, from our experiment, traditional machine learning approaches may not be optimal as a classification method for the current data.

Project description:We report here numerous novel genes and multiple new signatures which robustly predict prostate cancer (PC) recurrence. We extracted 696 differentially expressed genes relative to a reported PC signature from the TCGA dataset (n = 492) and built a 15-gene signature (SigMuc1NW) using Elastic-net with 10-fold cross-validation through analyzing their expressions at 1.5 standard deviation/SD below and 2 SD above a population mean. SigMuc1NW predicts biochemical recurrence (BCR) following surgery with 56.4% sensitivity, 72.6% specificity, and 63.24 median months disease free (MMDF) (P = 1.12e-12). The prediction accuracy is improved with the use of SigMuc1NW's cutpoint (P = 3e-15) and is further enhanced (sensitivity 67%, specificity 75.7%, MMDF 45.2, P = 0) when all 15 genes were analyzed through their cutpoints instead of their SDs. These genes individually associate with BCR using either SD or cutpoint as the cutoff points. Eight of 15 genes are individual risk factors after adjusting for age at diagnosis, Gleason score, surgical margin, and tumor stage. Eleven of 15 genes are novel to PC. SigMuc1NW discriminates BCR with time-dependent AUC (tAUC) values of 76.6% at 11.5 months (76.6%-11.5 m), 73.8%-22.3 m, 78.5%-32.1 m, and 76.4%-48.4 m. SigMuc1NW is correlated with adverse features of PC, high Gleason scores (odds ratio/OR 1.48, P < 2e-16), and advanced tumor stages (OR 1.33, P = 4.37e-13). SigMuc1NW remains an independent risk factor of BCR (HR 2.44, 95% CI 1.53-3.87, P = 1.62e-4) after adjusting for age at diagnosis, Gleason score, surgical margin, and tumor stage. In an independent PC (MSKCC) cohort (n = 140), these 15 genes were altered in PC vs normal tissue, metastatic PCs vs primary PCs, and recurrent PCs vs nonrecurrent PCs. Importantly, a 10-gene subsignature SigMuc1NW1 predicts BCR in MSKCC (P = 3.11e-15) and TCGA (P = 3.13e-12); SigMuc1NW1 discriminates BCR at 18.4 m with tAUC as 82.5%. Collectively, our analyses support SigMuc1NW as a novel and robust signature in predicting BCR of PC.

Project description:BackgroundGene signatures are important to represent the molecular changes in the disease genomes or the cells in specific conditions, and have been often used to separate samples into different groups for better research or clinical treatment. While many methods and applications have been available in literature, there still lack powerful ones that can take account of the complex data and detect the most informative signatures.MethodsIn this article, we present a new framework for identifying gene signatures using Pareto-optimal cluster size identification for RNA-seq data. We first performed pre-filtering steps and normalization, then utilized the empirical Bayes test in Limma package to identify the differentially expressed genes (DEGs). Next, we used a multi-objective optimization technique, "Multi-objective optimization for collecting cluster alternatives" (MOCCA in R package) on these DEGs to find Pareto-optimal cluster size, and then applied k-means clustering to the RNA-seq data based on the optimal cluster size. The best cluster was obtained through computing the average Spearman's Correlation Score among all the genes in pair-wise manner belonging to the module. The best cluster is treated as the signature for the respective disease or cellular condition.ResultsWe applied our framework to a cervical cancer RNA-seq dataset, which included 253 squamous cell carcinoma (SCC) samples and 22 adenocarcinoma (ADENO) samples. We identified a total of 582 DEGs by Limma analysis of SCC versus ADENO samples. Among them, 260 are up-regulated genes and 322 are down-regulated genes. Using MOCCA, we obtained seven Pareto-optimal clusters. The best cluster has a total of 35 DEGs consisting of all-upregulated genes. For validation, we ran PAMR (prediction analysis for microarrays) classifier on the selected best cluster, and assessed the classification performance. Our evaluation, measured by sensitivity, specificity, precision, and accuracy, showed high confidence.ConclusionsOur framework identified a multi-objective based cluster that is treated as a signature that can classify the disease and control group of samples with higher classification performance (accuracy 0.935) for the corresponding disease. Our method is useful to find signature for any RNA-seq or microarray data.

Dataset Information

Pareto Optimization Identifies Diverse Set of Phosphorylation Signatures Predicting Response to Treatment with Dasatinib.

Publications

Pareto Optimization Identifies Diverse Set of Phosphorylation Signatures Predicting Response to Treatment with Dasatinib.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets