Need advise regarding the use of univariate vs multivariate statistics

Dear stat experts

I have a dataset with 98 independent variables (or features in the multivariate machine learning case), 1 dependent variable (or target) with a N = 54.

Now I would like to explore which variables are most "assoicated" with the target variable. As I have a lot of features, I would like to sort out less important features.

I the univariate case I would first run a principal component analysis. Subsequently, I would run univariate regression...e.g. stepwise regression .
In the multivariate case I would first run a recursive feature elimination to get the most important features. Also here, I would finally run a univariate regression...e.g. stepwise regression on those important features.

Could someone please explain me the difference between those two approaches and when to use which one ?

Furthermore...and somehow related: Whats the difference between runnning a linear regression with the above-mentioned variables/features vs a regression using support vector regression (SVR) ?

Many, many thanks for your help!