Hello, With respect to the UCSC Genome Browser, the Table browser includes a standard linear correlation function. It may be helpful to review the exact methods this tool uses to perform the computation (as it is an example of correlation analysis accepted by other genomic scientists). http://genome.ucsc.edu/cgi-bin/hgTables. Click on the correlation button for details/methods.
Some considerations when performing this type of function on genomic datasets: 1) Learn how to distinguish between datasets that are sparse due to lack of experimental evidence vs a lack of actual biological significance/activity. The former may be inappropriate for many comparisons as a false "uncorrelated" result could occur. 2) Know that many of the better annotated/complete conclusion datasets are based on multiple underlying evidence layers. If you attempt to correlate between a conclusion dataset and one of it's input evidence layers or between two conclusion datasets that share an evidence layer, a false "correlated" result could occur. 3) Make efforts to remove datasets or portions of datasets that map to the genomic without specificity. Specifically, avoid known repeats or a false "correlated" result could occur, in particular if the the comparison is between short windows or at the base level. This question is very general and difficult to answer further without knowing the details of your experiment. A quick google search located this web site that discusses correlation and how to interpret the results for a variety of algorithms and data types. http://en.wikipedia.org/wiki/Correlation Jennifer Jackson UCSC Genome Bioinformatics Group Bogdan Tanasa wrote: > Hi all, > > I would like to ask a question about the correlations between 2 distinct > genome tracks : > let's say that there is a set of genome elements A and a set of genome > elements B. > > Would a correlation coefficient over an unified vector = vector A + vector B > be appropriate > to assess the statistical significance of the correlations of vector A with > vector B? thanks ! > > > -- bogdan > _______________________________________________ > Genome maillist - [email protected] > https://lists.soe.ucsc.edu/mailman/listinfo/genome > _______________________________________________ Genome maillist - [email protected] https://lists.soe.ucsc.edu/mailman/listinfo/genome
