Nadia Terranova (1), Konstantinos Ioannou (1, 2), Pascal Girard (1), Alain Munafo (1)
(1) Merck Institute for Pharmacometrics, Merck Serono S.A., Lausanne, Switzerland; (2) School of Computer and Communication Science, EPFL, Lausanne, Switzerland.
Objectives: Clinical models of tumor dynamics generally omit information on individual target lesions (iTLs), and use the total tumor size (TS) as a continuous variable to model the tumor time-course. However, differences in lesion dynamics might be predictive of tumor progression. To exploit this information, we have integrated knowledge from signal processing and machine learning into a novel and flexible approach for the non-parametric analysis of iTLs [1, 2]. We called this new methodology ClassIfication Clustering of Individual Lesions (CICIL). In this work, we present the CICIL tool, a Java-based cross-platform implementation of the CICIL methodology, recently made available to the scientific community [2].
Methods: The CICIL methodology relies on the classification of iTLs based on functional and anatomical criteria, and it consists on a workflow accommodating the assessment of similarity among dynamics of lesions classified as belonging to the same anatomical site (intra-class analysis) or to different sites (inter-class analysis). Such degree of similarity is assessed through cross-correlation measures, and the interpretation of the results is facilitated by the k-means clustering [2].
To enable the efficient execution of this methodology and to assist the interpretation and visualization of each individual step in the workflow, CICIL has also been implemented in a user-friendly Java-based framework [2]. The CICIL tool, through its functional and interactive graphical user interface (GUI), enables a user to seamlessly create new projects, import and manipulate datasets, and run the CICIL workflow to obtain a series of informative graphical plots and well-structured statistical summaries. Moreover, the tool is modular and flexible as it provides a high degree of customization for its core components. For example, the iTLs classification can be defined by using standard terms automatically extracted from the dataset through a text-mining algorithm or a set of keywords directly defined by the user. Similarly, the user can select the desired results that she/he wants to export and automatically generate customized reports directly through the GUI.
Results: The CICIL tool’s executable (JAR file) is publicly available as Supplementary Material of Terranova et al. [2] along with a use case based on a mock dataset. The tool can be executed on operating systems which contain a version of the Java Runtime Environment, minimum v1.7, and has been tested in Windows 7 and 8. System requirements and application features are described in the respective user guide embedded in the tool.
Conclusions: The CICIL tool constitutes a user-friendly and flexible platform enabling a straightforward execution of the CICIL methodology to efficiently analyze and understand large-scale datasets prior to modeling. The results can then guide the modeler in determining whether a total TS evaluation might reasonably predict tumor lesion behavior, or potential differences in responses, within or across tumor site classes, should be taken into account for a particular case study and for the questions to be addressed.
References:
[1] N. Terranova, P. Girard, U. Klinkhardt, A. Munafo. Analysis of individual target lesions for tumor size models of drug resistance: a new methodology encompassing signal processing and machine learning. ISSN 1871-6032. PAGE 24 (2015) Abstr 3399.
[2] N. Terranova, P. Girard, K. Ioannou, U. Klinkhardt, A. Munafo. Assessing similarity among individual tumor size lesion dynamics: The CICIL methodology. CPT Pharmacometrics Syst. Pharmacol (2018).
Reference: PAGE 27 (2018) Abstr 8497 [www.page-meeting.org/?abstract=8497]
Poster: Methodology - Other topics