Hi all,
I am posting because I have a doubt about some work I am doing with microarray (but it works with almost any omics data) analysis. More in detail I am not a biologist but an engineer and I like to work with algorithms. I developed an algorithm for classification which works well and got comething published. Now I have added variations that allow including biological knowledge and a hierarchical data representation, and I have new results that I don't know if they are interesting enough to be submitted (at least at a conference): analizing public data (MAQC) what I get is the same mean predictive power but basically half variance in the error rate over repeatd simulations. analyzing some private data (I cannnot make them public cause are not mine) I got sensible differences analyzing the gene lists with GSEA or biograph and other tools. The differences state how the new method finds constanly genes more related to leukemia than the first one.
I didnìt do any enrichment analysis with public data because I don't know what to look for while in the private data I know they are about leukemia... even so I am waiting the hospital guys to evaluate my results but it looks like they will take some 4 5 months before answering.
What do you think about communicating these results? do you thing they are worth an attempt? Furthermore.... do you know some easy tool to evaluate the quality of gene lists related to a subject... this may help in extracting additional results hopefully stating that the new method is better :p
thank you very much Mattia
Thanks for the reply I know that it is not possible to judge the work itself without any description, I try to reformulate the question from a more general point of view: Would it be interesting a work about a classification algorithm with predictive accuracy comparable to state of the art results and with improved (w.r.t. previous implementation) biological siginificance of the proposed gene lists ? would it gain some interest if the proposed gene lists are also compared to external already known results in terms of significant genes?
About evaluating the quality of a gene list I found that it is not easy at all and totally agree with you.. that's why I wanted some advie and I'll probably end up asking some expert's help in gene list evaluation even if it means wait months for an answer. thanks again for the time