Bioinformatics
The PCPGM Microarray team performs standard validation procedures for the quality of RNA samples, labeling, and hybridization for all microarray studies. Our microarray instruments are on service contract and subject to routine PM maintenance under those contracts. In addition, there is an examination of the data output from either Affymetrix or Illumina expression analysis platforms. Before any analysis is performed the data are normalized based on results from our examination of the output and QC process.
Affymetrix
- We use GenePattern1 for analyzing Affymetrix data. Briefly we do MAS5.0 normalization on intensity values scaled to 500. We then do comparative marker selection and analysis and generate fold-change and scores for each of the probes. We provide standard description (encrypted with various tags that can be extracted) for each probe but do not provide detailed annotation such as Gene Ontology.
Illumina
- We use BeadStudio (Illumina, Inc.) to perform the analysis. Briefly we do background correction, average normalization, and compare control vs experimental sets.
Affymetrix
- We provide fold-change and a score for each of the differentially expressed probe sets. The top 200 up and down regulated genes are separated and presented as an html file with suffix “selected”. A heat-map with probes linking to Affy site (www.affymetrix.com) is also generated for each comparison made.
Illumina
- We provide fold-change and p-values for each of the differentially expressed probes with an experimental set. This is obtained for each comparison made between groups of samples.
More complex types of analysis such as Gene Set Enrichment Analysis (GSEA), customized heat maps, pathway analysis via Ingenuity, and additional simple analysis plots using R may be possible but we consider this sort of analysis beyond the scope of our "Full Service" and these "advanced" services would be at an additional charge over the base service cost. Furthermore, the experimental design and statistical power of the replicates provided would also need to be robust to justify the use of these tools to get further useful results. If you are interested in GSEA and others of these tools as an added feature, please contact us to discuss this application. Projects submitted under “Full Service” will have their data analysis performed and a primary analysis package returned to their GIGPAD account data page in the standard report formats as described.
NOTE:
In both Affymetrix and Illumina analysis, comparison between groups can be made with statistical significance ONLY when each group has at least three biological replicates. If less that three replicates are used statistical significance cannot be established and the customer has to primarily use fold-change to obtain differentially expressed genes.
1. Reich M, Liefeld T, Gould J, Lerner J, Tamayo P, Mesirov JP: GenePattern 2.0, Nat Genet 2006, 38:500-501
These
links are offered as options where you may wish to take your analysis data for GO and other post analysis annotation input.
Report turnover time is directly proportional to the complexity of analysis and number of projects in the queue. Additional analyses may be offered on a case-by-case basis. The scope of each is evaluated based on the fit to our staff resources. If we find the type of analysis you are seeking is beyond the scope of analysis we are able to provide, we will suggest contact with others that may be interested in collaboration.
- Why do we recommend BIOLOGIAL REPLICATES?
- The probability of successfully detecting a differentially expressed gene depends on three main factors. The true magnitude of the differential expression, magnitude of random variation (background noise), and the number of replications.
- Biological replicates are RNA from different individual cultures, organisms, etc collected across a set of conditions defined for your experiment but all are subjected to the same conditions with the exception of the applications of control and experimental parameters.
- A minimum of three (3) biological replicates is recommended. Appropriate numbers of biological replicates help increase the statistical power of the analysis, allow for resolution of any outliers, and lend more robustness to interpretations of data detected by the experimen. We recognize controlling costs is an essential element for investigators as it is for us. However, weighing the informativeness of the data vs. the cost of results should also be a prominent factor in deciding on the design of a project. Perhaps fewer conditions with more replicates may be an acceptable strategy to get the most out of a project.
- Biological replicates differ from technical replicates in that a technical replicate is simply taking samples from one source tube and running it multiple times. With the commercial arrays manufactured over the last several years, variance from array to array is very low now. Therefore, the technical replicate does not add much value to analysis and we do not generally recommend them to be included.
- Pooling samples together is not recommended. If you did find a need to pool samples, this does not replace the need for replicates.
