Is it proposed to use replicates (and if yes, how many) when doing spatial omics, using the same type of tissues but from different animals within the same phylum?
Your study design depends on the question you are asking, the availability of biological materials, and other constraints like cost.
If the animals are different species, I would consider those independent data points and not "replicates".
Spatial omics - as mostly all omics - is largely hypothesis-generating. If you like to test omics-sepcific hypotheses, the huge difficulty is to formulate the overarching (statistical) model that contains the meaningful hypothesis within a meaningful parameter space. That's usually all not clear at all - but this would be required to get a hold of how (biological) replicates can be integrated (into this same statistical model), and how many would be required for a reasonable power (which neccesarily refers to the hypothesis you want to test and gainst what alternative in the parameter space you want to test it). As far as I can see, the parameter space is always multi-dimensional (often with hundred or even thausands of dimensions), what makes it difficult to define (meaningful) alternatives. When replicates are used in omics experiments, they mainly serve the purpose to test individual components (genes, proteins, metabolites...) and not the structure, pattern, or profile. This is actually not an omics approach but rather splitting the omics data into statistically unrelated problems in one-dimensional parameter spaces (where simple hypotheses can be tested using known old procedures).
Thank you both for your replies. TO be more exact, i will proceed with the Visium spatial transcriptomics technique. Indeed, the aim is to do hypothesis generation. I am just wondering since all the related papers do not say anything about replicates (obviously the cost plays an important role) what is better to do.. Different data points have been considered.
Sure, costs are relevant, but much more relevant is the clear definition of a statistical model that maps meaningful biological features. This is simply not really clear how to do that in a multidemsional space.
If you found some kind of pattern (however you define it), it is certainly good practice to repeat the entire experiment and analysis at least onece (better twice or trice) to see if the pattern you identified occurs (more or less) robustly. If the observed pattern indicates some biological interpretation, you can go back to biology ind infer the hypothesized effects by new experiments (knock-ins, knock-outs, blocking, competeing, histology, etc.).