Cutoff Finder: a comprehensive and straightforward Web application enabling rapid biomarker cutoff optimization.

Budczies J, Klauschen F, Sinn BV, Győrffy B, Schmitt WD, Darb-Esfahani S, Denkert C. · 2012

Read Cutoff Finder: a comprehensive and straightforward Web application enabling rapid biomarker cutoff optimization. by Budczies J, Klauschen F, Sinn BV, Győrffy B, Schmitt WD, Darb-Esfahani S, Denkert C. (2012) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc3522617.txt.

Introduction

A biomarker becomes a treatment decision only at a cutoff value, and Jan Budczies and seven colleagues built Cutoff Finder to choose that value well in 2012. Their free web application bundles optimization and visualization methods where no standard tool existed.

Users upload molecular data, with outcome or survival data if available, pick a method, and receive the optimal cutoff plus publication-ready plots. Some methods use only the marker's own distribution; others tune the split against outcome or survival. A breast cancer demonstration analyzes estrogen and progesterone receptor expression against lab-determined status and metastasis-free survival. Choosing the best cutoff remains partly judgment, the authors admit, but the search is now systematic.

Catalog record

CREATORBudczies J, Klauschen F, Sinn BV, Győrffy B, Schmitt WD, Darb-Esfahani S, Denkert C.
YEAR2012
COLLECTIONPAPERS
FORMATTXT
FILE SIZE32 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC3522617

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Cutoff Finder: A Comprehensive and Straightforward Web Application Enabling Rapid Biomarker Cutoff Optimization By Jan Budczies, Frederick Klauschen, Bruno V Sinn, Balázs Győrffy, Wolfgang D Schmitt, Silvia Darb-Esfahani, Carsten Denkert, Paul van Diest Published in: PLoS ONE DOI: 10.1371/journal.pone.0051862 ABSTRACT Abstract Gene or protein expression data are usually represented by metric or at least ordinal variables. In order to translate a continuous variable into a clinical decision, it is necessary to determine a cutoff point and to stratify patients into two groups each requiring a different kind of treatment. Currently, there is no standard method or standard software for biomarker cutoff determination. Therefore, we developed Cutoff Finder, a bundle of optimization and visualization methods for cutoff determination that is accessible online. While one of the methods for cutoff optimization is based solely on the distribution of the marker under investigation, other methods optimize the correlation of the dichotomization with respect to an outcome or survival variable. We illustrate the functionality of Cutoff Finder by the analysis of the gene expression of estrogen receptor (ER) and progesterone receptor (PgR) in breast cancer tissues. This distribution of these important markers is analyzed and correlated with immunohistologically determined ER status and distant metastasis free survival. Cutoff Finder is expected to fill a relevant gap in the available biometric software repertoire and will enable faster optimization of new diagnostic biomarkers. The tool can be accessed at http://molpath.charite.de/cutoff. INTRODUCTION The use of diagnostic, prognostic and predictive biomarkers is of increasing importance in many areas of medicine. Many markers are measured in laboratory assay as continuous variables or as ordinal variables. Gene expression is usually measured on a metric scale, using qRT-PCR or hybridization based methods, for example using single gene assays such as TaqMan or on a global scale using DNA microarrays. In tissue-based diagnostics, protein expression is usually evaluated by immunohistochemistry (IHC) and quantified on an ordinal scale using the percentage of stained cells, staining intensity or combinations of these. Examples of such combinations include the Allred score that takes values between 0 and 8 [1], [2] and the immunoreactive score (IRS) that takes values between 0 and 12 [3], [4]. In order to translate a continuous or ordinal diagnostic variable into a clinical decision, it is necessary to determine a cutoff point and to stratify patients into distinct groups each requiring a different kind of treatment [5]. Although the mean or median value of the diagnostic factor was used before, it is often desirable to determine cutoff points based on the distribution of the variable or by optimizing the correlation with clinical outcome or response to a treatment. Methods for cutoff optimization include minimization of p-values or maximization of combinations of test sensitivity and test specificity [6]. X-Tile is a freely available tool that is specialized to the analysis of survival data and uses a minimal p-value approach for cutoff optimization [7]. Previously, like many others, we manually investigated and optimized cutoff points of molecular markers [8]–[10]. However, methods for cut-off determination vary among published studies and the underlying algorithms remain obscure in many instances. The literature is filled with newly identified biomarkers, but only a small minority has found acceptance in patient care so far. This is in part due to overestimation of the true effects in the first study and a low reproducibility in subsequent validation studies. Cutoff optimization was demonstrated to contribute to overestimation of results when multiple cutoff points are investigated and the problem of multiple hypothesis testing is ignored [11], [12]. Conflicting the need of a higher degree of objectivity [13], cut-off point determination is often done in a non-systematic manner and therefore among the causes for a poor reproducibility of biomarker studies. The reason for this is the lack of comprehensive and easy-to-use tools for cutoff determination. Therefore, we developed Cutoff Finder, a bundle of optimization and visualization methods that are accessible via the internet (http://molpath.charite.de/cutoff). The purpose of the web application is twofold: (i) As not a single methods can be considered to be optimal for all kind of data and all clinical situations, we implemented a multitude of five methods for cutoff determination. (ii) To avoid overestimation of the true effect, the robustness when varying the cutoff point can be investigated in overview plots showing effect sizes with confidence intervals. The workflow through the web application is shown in Fig. 1. [Figure: Workflow of Cutoff Finder.The track below the icons refers to the places where the steps of data processing are done. These are the steps of data procession: 1. Data are uploaded from a tab-separated file or imported from one of three example data sets. 2. The user selects the biomarker and optionally outcome and survival variables from the table columns. 3. The user selects the method for cutoff determination. 4. The user chooses the set of plots to be generated. 5. The optimal cutoff point is determined and analysis plots are generated using R as statistical engine. 6. Cutoff point and plots are shown at the results webpage.] METHODS Implementation and Availability Cutoff Finder is a freely available web application that can be accessed using an arbitrary web browser (http://molpath.charite.de/cutoff). The web application is implemented as Java Server Pages (JSPs) that connect to R using the R package and TCP/IP server Rserve [14]. The R statistical language [15] serves as an engine for all statistical computing and visualization. The R code for cutoff optimization and the plot functions can be found in Text S1. The R file is loaded as static source to the R server. For each cutoff analysis, the web application calls the wrapper R function get.cutoff() that serves as controller. The controller calls specialized R functions for each kind of cutoff optimization and for each kind of plot. Results of the cutoff optimization and plots are returned to the web server. Using the R code, results and plots of the web page can be reproduced 100%. Analysis Workflow The steps of data procession are shown in Figure 1: - The data are uploaded from a tab-separated table, rows representing patients and columns representing variables. The maximum number of rows is 5000, the maximum number of columns is 50. - The user selects the biomarker and optionally outcome and survival variables from the table columns. - The user selects the method for cutoff determination. - The user chooses the set of plots to be generated. - The inquiry is send to the statistical engine R. The optimal cutoff point is determined and analysis plots are generated. - The user is directed to the results webpage, where the optimal cutoff point and plots are shown. Cutoff Determination A multitude of five methods is offered for cutoff optimization. The first method is based solely on the distribution of the biomarker in the sample cohort. Methods 2–4 deal with optimization of the correlation with a binary variable, for example response to treatment. The last method is devoted to optimization of the correlation with survival. - Fit of mixture model: A mixture model of two Gaussian distributions is fitted to the histogram of the biomarker. This procedure is implemented using the function flexmix from the R package flexmix [16]. The optimal cutoff is determined as

Read the complete text