Performing PCA on count data with many true 0´s [duplicate]

Question

I have a dataset with behavioural observations that are split into different types within each category.

For example one category would be: "Boldness". Within "Boldness" 7 different types of behaviour exist, which would be my different variables (coloumns). For each of these variables I have count data, which often are true 0´s. The rows are repeated measures for different individuals.

I want to reduce my data to one final score per category (e.g., "Boldness"). In the literature, I found researchers conducting a PCA and using the unit-scale loadings of the first component as such as score.

However, my problem is:

1. The distribution of my data (poisson)
2. The true 0´s in my data (I believe)

Below an example when I tried to box-cox transform the data:

Any help on how to achieve this potential final score/how to conduct a PCA with this kind of data in R, would be massively appreciated. Maybe this is even ok already to continue?