Matched molecular pair analysis (MMPA) and matched molecular series. Finds every pair of molecules differing by one structural change, then aggregates across pairs into evidence-backed transformation rules with N, mean and median delta, standard deviation and fraction improved. Also extracts matched series — one context with three or more substituents explored. Use when deciding which substituent change to make next, or mining a dataset for transformations that reliably move potency. Requires curated log-scale activity. Does not need a shared core, unlike rgroup-sar and free-wilson.
