Integration of multivariate data in systems biology
Sahar Hassani · BIBSYS Brage (BIBSYS (Norway)) · 2012
Owing to the rapid rate of development in the field of systems biology researchers have faced many new challenges with regard to handling the large amount of generated data sets originating from different –omics techniques, integrating and analyzing them and finally interpreting the results in a meaningful way. Different statistical methods have been implemented in the field of systems biology. The use of chemometrics approaches for the integration and analysis of systems biology data has recently increased. Different chemometrics methods are potentially available for integrating –omics data and detecting variable and sample patterns. An important challenge is to decide which method to use for the analysis of –omics data sets and how to pre-process the data sets for this purpose. Special attention needs to be given to the validity of the detected patterns. In this study we have been working on developing multi-block methods for integrating different types of systems biology data and investigating the co-variation patterns among the measured variables. A special focus was given to the validation of the results of the multi-block methods CPCA and MBPLSR. Different types of graphical tools were introduced for the purpose of validation. We have also developed pre-processing techniques that could explicitly be used for lipidomics data sets. A framework was built for pre-processing, integrating, analyzing and interpreting the lipidomics data sets. The framework was then used for the analysis of a lipidomics data set from a human intervention study. Working on the development of the validation tools required an understanding of the concept of DFs consumption during the multi-block modeling. Therefore, we ran simulation studies where we investigated the number of DFs that were consumed during the modeling processes of PCA and CPCA. Another important issue for applying multi-block methods is the choice of the deflation method. Hence, we studied different deflation strategies available for Multi-block PCA and investigated their interpretational aspects.