Closed AlexanderPico closed 5 years ago
I think since this covers enrichment analysis we should also cover the statistical consequences of this. What I have in mind is: Indicate that pathway analysis can be seen as an extension of omics statistics analysis and can be a knowledge-based form of false discovery reduction Indicate that technologies used highly influence statistical sensitivity and that in general low expressed/present entities will have higher errors (and that the interaction of biology and technology may affect what is measured as more present (think CG rich areas and hairpins on arrays, there are plenty of examples from other techniques too. Discuss how biology (for instance multiple parallel routes/isoenzymes for some steps, completeness and even size of pathways may affect statistics Based on the above, discuss the limited value of more advanced statistical procedures since you cannot get further than detecting pathways that you should seriously look at anyway. Then still discuss basic enrichments methods (the basic Z-score algorithm present in PathVisio, GSEA, the pruning approaches in ontology enrichment (using GO_Elite) to discover things not possibly covered in pathways. By actually running these tools we hope to also uncover what is so hard about some of these. I typically see people move not to what they really need, but to what they can get to run quickly (GO_Elite being on the side what of what is useful but not often used). I like the idea of not just doing data overlays but also interactions in Cytoscape. But that will make it quite full and means we need to plan well. I have seen in the past that letting people work at there own speed helps. More experienced people then can skip some parts and go to what is new for them.