For the tasks you describe, I use perl and in particular the Ensembl perl API if the genome is annotated in Ensembl. Otherwise, I use R for statistics, machine learning and some computational biology. I had to go with R because what I needed a few years ago wasn't available in python or perl. My impression is that nowadays, python can do a decent bit of both bioinformatics and machine learning (but maybe not so much statistics and computational biology compared to R). If you know one well, you won't need the other for most of what you need to do. So the choice should be to go for what you think will take you furthest for the least amount of time invested.