DNAVision uses a variety of statistic and bioinformatic tools to analyse small, large and/or complex datasets obtained from multiple microarrays.
By combining different methods, we provide customers with meaningful, reliable data related to biological phenomena or pathways. We also analyse data generated from different platforms, using a variety of normalisation methods (median normalisation, rma, gcrma .).
Technical brochure: Bioinformatics at DNAVision
DNAVision provides you with a set of tests to asses the quality of your data. These tests are not only based on the manufacturers requirements (Affymetrix’s standards) but also on the MicroArray Quality Control (MAQC) guidelines in accordance with the ISO17025.
We develop our own tools (to be released soon) that do follow these guidelines for optimal and reproducible quality analyses. We perform reproducibility analysis of reference RNAs on a regular basis and show extremely high reproducibility, in accordance with the MAQC phase 1 datasets.
DNAVision provides you with several clustering methods (hierarchical clustering, self-organizing maps, k-means clustering, multi-dimensional scaling,etc.) to uncover patterns of sample or gene expression data and the relationships between them.
Genes and their expression patterns can be visually characterized based on their ontology, cell or tissue location and more precisely within a cellular pathway. These methods can predict the genes associated with biological phenomena or discrete steps in the pathway of interest. DNAVision can analyze any pathway of interest.
DNAVision also provides a variety of tools to answer specific needs within complex data sets. These include classic t-test (paired or unpaired, randomised or not,etc.), correction for multiple tests (e.g., permutations, Benjamini-Hochberg correction) and analysis of variance (ANOVA), to reliably identify differentially expressed genes.
In addition, for large volumes of data, we use class prediction tools to identify genes capable of discriminating between experimental parameters or sample phenotypes. Identification of key genes can be cross validated using mathematical methods such as K-nearest neighbor (with k=1 or 3), compound covariate predictor or support vector machines for their classification performance.