The idea that one could look at a person's genome and make realistic predictions about their health prospects and the likelihood any of a vast array of different diseases was really sold in the early days of genome sequencing, but I always scoffed at the idea. Sure there were cases when you could do it for specific inherited mutations and their impact on some diseases, but there were some really wild statements about predictive powers that are obviously not coming true. It is very useful to type tumors and compare them to somatic tissue fro the same patient, but the heterogeneity of tumors complicates this.
With regard to base pair changes, whether you call them SNPs because they fall below some arbitrary threshold, or rare variants, make no difference. One needs a secondary terminology to describe the effect on protein function or activity. There are examples of single base changes which keep the primary amino acid sequence the same, yet alter the protein activity.
