There's likely significant individual variation in bioavailability. I would start with 2-5K/day, then measure and iterate.