How do you find similar gene expression pattern to your GOI?

Question

[Hushus]

Hey all,

I have a set of expression profiles throughout embryogenesis (Drosophila Flybase). These have quantitative values associated with them at 0-2hr increments. I also have a particular gene of interest.

I would like to find a set of genes that share the same gene expression PATTERN (not quantity) at a specific time span. As in, while this genes expression steadily increases from time A to B then decreases from to B to C, these list of genes do as well.

Is there a tool to do this??

[Jake Warner]

Maybe you could cluster the data, then find which cluster your GOI is in. The other cluster members should have a similar profiles.

A more complex way is to generate clusters using your input genes as initialization centroids. Although maybe somebody could give some insight because I have no experience with this.

Answer 1

[Jean-Karim Heriche]

Compute similarities between the time series of interest and all the others then simply rank them in decreasing order. When plotting them you might notice an elbow that would indicate a convenient threshold or use any kind of threshold that's practical, e.g. if you can only do experimental validation on 5 genes then just take the top 5. The trick would be to use a good similarity measure. If you don't expect noise in the timing (i.e. all data are perfectly aligned in time) then something like the cosine similarity should do the job. However, if you suspect that there may be some slight shift in timing (e.g. some embryos might progress though some stages a bit faster than others), you should consider a similarity measure based on dynamic time warping.