Francesco Gadaleta discusses using penalized regression methods like the lasso and elastic net to select covariates and build gene networks from microarray experiments. These methods allow for variable selection and network construction while accounting for issues like multicollinearity. Gadaleta demonstrates his "Select and Connect" approach on both synthetic data from a known regulatory network as well as real gene expression data, showing it can accurately identify hub genes and network connections. He also discusses parallelizing the computations and integrating multiple networks.