Checking normality of a variable

Question

I want to perform linear regression with interaction between 2 factors for 80 variables (human data, i.e., age, weight, height, hemoglobin etc.). I checked for normality of variables and more of them had failed shapiro.test.

Is it statistically correct to transform (log transform) and perform the test. or do I seperate them group-wise and try shapiro again?

Thank you