Abstract— This paper presents a user-friendly web-based collaborative environment for analyzing, assessing the quality of large multi-level clinical datasets and deriving predictive models. The Multi-Modal Medical Data Analysis Platform (3MDAP) follows two main objectives: a) to empower the user to analyze with ease clinic-genomic data in order to get simple statistics on selected parameters, perform survival analyses, compare regiments in selected cohort of patient and obtain genomic analysis results, and b) to perform heterogeneous clinical data modeling for deriving and cross-validating in multiple datasets predictive clinic-genomic models of patient response, and assessing the value of candidate biomarkers. 3MDAP’s enhanced functionality is coupled with a security framework for enabling user authentication and authorization, a set of services that facilitate the process of loading and retrieving data from a data-warehouse (either locally based or in a cloud), and a widget-based front-end environment for assisting users in interacting with the platform’s functionality in a user friendly manner. For each running analysis, 3MDAP supports an engine to create dynamically analysis reports. Last, the framework provides an internal database where a full analysis record of an executed analysis is stored, including metadata information (i.e. timestamp information, the examined data, any memory constraints, the dynamically generated reports in both .pdf and .html format, and etc.) in order to be used for future reference. I. INTRODUCTION The proposed platform for cancer clinic-genomic data analysis is centered in empowering the user (e.g. clinical researcher or bioinformatician) to obtain simple descriptive statistics and compute with simple, high-level operations predictive models that can be seamlessly validated in multiple datasets within a single platform in the same session. The main design objective is to allow any user to use the platform even without expertise in computational tools such as the R software environment for statistical computing [1]. The developed predictive analysis functionality featuring a comprehensive clinical trial data-viewer has been largely driven by the clinical scenarios for the INTEGRATE VPH project [2] as well as by extensive discussions with experts bioinformaticians/clinicians involved in the project. 3MDAP is developed by the Computational Medicine Laboratory (CML) of the Institute of Computer Science (ICS) in the Foundation for Research & Technology - Hellas (FORTH), Vassilika Vouton, P.O Box 1385, GR-71110 Heraklion, Crete, Greece G.C. Manikis (phone: +30-2810-391593, e-mail: gmanikis@ics.forth.gr). E. Maniadi (phone: +30-2810-391479, e-mail: maniadi@ics.forth.gr). M. Tsiknakis (phone: +30-2810-391690, e-mail: tsiknaki@ics.forth.gr). K. Marias, (phone: +30-2810-391672, e-mail: kmarias@ics.forth.gr). *G.C. Manikis and E. Maniadi contributed equally to this work. This paper presents in detail the platform explaining how it is capable to enable scientists from diverse backgrounds to employ with ease (at the push of a button) a) sophisticated statistical analysis tools that play and important role in deeply understanding and preparing the available multi-level data for further analysis, and, b) to derive predictive models (again at the push of the button) from cancer clinical trial data. II. 3MDAP PLATFORM A. System Architecture The idea behind 3MDAP is to provide users with a web- based interface that supports user authentication and authorization, data handling, execution of the tools and models, and visualization and storage of the analysis reports. To achieve this goal, the programming aspects of the different environments and languages adopted for implementing the framework's facilities, and the connectivity process which allows the interaction between these components are kept at the back-end of the framework, hiding the complexities of the computational infrastructure. The architecture and specifications of the developed framework are divided into the following fields: • The core functionality of the platform • The authentication and authorization process • The data retrieval system • The web services infrastructure 1) The core functionality From the technical perspective, the core functionality is composed of the front-end and the back-end component of the platform. The front-end, hiding the complex infrastructure is based on the Liferay Portal [3]. Liferay Portal is an enterprise web framework based on Java technologies. Our Liferay based front-end is enhanced with JavaServer Faces (JSF), a Java technology for building component-based user interfaces for web applications. An Ajax-based JSF framework named as PrimeFaces [4] was chosen to be used in 3MDAP, offering over 100 individual components, covering a diverse range of widgets including Ajax, input fields, buttons, data display controls, panels, overlays, menus, charts, dialogs, multimedia presentations, drag/drop and other controls. The back-end (Fig.1) consists of a complex heterogeneous environment of several software components. The main part of it is the statistical and predictive modelling analysis software scripts, implemented in R language [1] and using publicly available libraries from its large repository. To facilitate embedding R functionality in our Java-based interface, a client/server concept using TCP/IP protocol [5] is used for the communication between the R system and the Multi-Modal Medical Data Analysis Platform (3MDAP) for analysis and predictive modelling of cancer trial data Georgios C. Manikis*, Evangelia Maniadi*, Manolis Tsiknakis, and Kostas Marias, Member, IEEE