There have been several large studies of identical twins separated at birth. For a given twin, the strongest correlate of IQ is the other twin's IQ. Likewise, the non-identical but biologically-related siblings do have moderately correlated IQs. However, there is barely any correlation between a twin and their adoptive parents or siblings. This suggests that genetics and prenatal environment determine most of IQ.
There is also strong evidence from gene sequencing that genetics directly affects IQ. Genes that affect sphingolipids, chemicals important in the brain, are known to affect IQ. A variation in the SNAP-25 gene has been found to increase intelligence by an average of 3 IQ points.
And while it has become a cliche, it is still true that correlation is not causation: coming up with correlation is easy, but solid scientific works need to prove that the correlation is not spurious. Most studies don't (this is sadly not specific to discussions about IQ, and is a plague of many studies in too many areas - good statistics are hard).
Yes, and I am closely acquainted, through attending their graduate student "journal club,"
http://www.psych.umn.edu/courses/fall10/psy8935/default.htm
with most of the researchers who have conducted the best of those studies. An earlier reply of mine here on HN
http://news.ycombinator.com/item?id=2696680
corrects (with citations to current research publications) some of the overstatements seen in popular literature about what twin studies show. There is also not any single gene variant that has large effect on IQ, in the view of those researchers about what the most up-to-date gene association studies show.
You're leaping across logical gaps here. IQ can be predicted by a number of factors (parental wealth, IQ, and education are a few), and research on intelligence is fraught with identification problems. It's hard to believe claims of a "genetic underclass."