Current feed energy and protein values in Indonesia is based on total digestible nutrient (TDN) and crude protein (CP), respectively. This system is rather old and other countries have moved towards a new system, in which feed energy and protein are represented by metabolizable energy (ME) and metabolizable protein (MP), respectively. This study aimed to perform an empirical modelling to estimate feed ME and MP contents by their chemical composition. A total of 134 feeds from various categories (dry forage, fresh forage, silage, concentrate rich in energy, concentrate rich in protein and by-product) were integrated into a database. Values of TDN and CP were regressed against ME and MP, respectively. The value of ME was predicted from neutral detergent fiber (NDF), non-fiber carbohydrate (NFC), ether extract (EE) and CP whereas MP was predicted from rumen degradable protein (RDP) and rumen undegradable protein (RUP). The RDP to RUP ratio was regressed to MP in order to obtain optimum value of the ratio. Results showed that TDN and CP could predict quite accurately ME and MP by explaining 78.2% and 92.7% of their total variations, respectively. The ME was accurately predicted by NFC, NDF, EE and CP, whereas MP was accurately predicted by RDP and RUP. Lower RDP/RUP led to a higher MP percentage to CP. It can be concluded that TDN, CP and other feed chemical constituents could predict quite accurately ME and MP. This indicates that the old system is still valid to be used for formulating ruminant diets.