Hi;Hope someone can answer my question.The vanna volga (The vanna-volga method for implied volatilitilies, A.Castagna, F.Mercurio) method requires just 3 inputs to form a volatility smile.ATM and 25C/25P imp. vols. This is not a problem if the quotes are readily available from the market, but not so for illiquid fx crosses...The intention is to build a vol smile for an illiquid FX cross.The article by U. Wystup "Quanto Options" derives the currency triangle formula, analogous to the cosine law formula.It is only valid for ATM vols. For illiquid Fx crosses, the ATM vols can be inferred from the currency triangle formula , using more liquid components of the fx pair.ie for clarity, assume MYRKRW is the illiquid cross,use USDMYR and USDKRW to imply the ATM vol of MYRKRW.However if one were considering smile effects , how would the the non-ATM cross vols be calculated/implied/inferred? i.ehow would the 25DCall imp. vols for MYRKRW (quotes not available in the market) be related to25DCall of USDMYR & USDKRW (which are readily available in the market)?&25DPut imp. vols for MYRKRW (quotes not available in the market) be related to25DPut of USDMYR & USDKRW (which are readily available in the market)?any other methodologies, references to articles to infer the non-ATM call/puts of the fx cross from the its constituents would be much appreciated