N of 1 study looking at complex phenotypes and SNPs


New Member
Hi All, I have a complicated question and can't find a resource that answers it comprehensively. I'm designing a study looking for linkage between a complex phenotype; how people respond to various different topical treatments. This will not be published in a scientific journal so will be relying on anecdotal information from patients on several factors we put together such as rating of the product on various criteria, self administered tests, etc. Basically, collecting various anecdotal evidence but will be lacking in empirical data. We may be incorporating sensors that talk to an app but again these will be self administered and not stringently controlled.

Here is the study design. I would appreciate and feedback and would love to know if it is worth pursuing:

1) obtain gene expression of treatments on a cell line to obtain a list of candidate genes.
2) pick several gene candidates and sequence SNP's known to have high frequency in the population. The genes are in pathways known to be involved in the phenotype of interest.
3) have people obtain baseline measurements of interest before the 'study' begins that will act as a control.
4) provide samples of the different treatments randomly to the population and monitor progress based on list of criteria we are following.
5) run a linear regression or ANOVA on each independent variable compared to dependent variable to screen for linkage.
6) run a complex linear regression/MANOVA on the variables that have a low p value.

The problem is there are so many factors that effect the outcome. However, I'm thinking that using the N of 1 study design will help account for some of these factors. There are also many pathways involved in the phenotype of interest but due to budget constraints, our strategy is to tackle each pathway in a linear order, starting with one pathway making our way down the line.

My questions are:
1) Is this approach feasible? We don't need hugely robust data some linkage of interest that we can then re-evaluate in a more robust way with a subset of people.
2) Is there a better way to analyze the data, a simpler approach, than linear regressions or MANOVA analysis?

I appreciate any feedback. I'm more of a biologist than a biostatistician so may be completely off my rocker here:)
Last edited: