The present invention pertains to the field of toxicological assessments for risk stratification of chemical compounds. Specifically, it relates to a method for diagnosing liver toxicity. It also relates to a method of determining whether a compound is capable of inducing such liver toxicity in a subject and to a method of identifying a drug for treating liver toxicity. Furthermore, the present invention relates to a data collection comprising characteristic values of at least five metabolites, a data storage medium comprising said data collection, and a system and a device for diagnosing liver toxicity. Finally, the present invention pertains to the use of a group of metabolites or means for the determination thereof for the manufacture of a diagnostic device or composition for diagnosing liver toxicity in a subject. For each sex, a different metabolome pattern, i.e. a different set of analytes is disclosed. The liver toxicity markers are mainly selected from free fatty acids, but also include various phosphatidylcholines, Hydroxyphenylpyruvic acid, alpha-Tocopherol, Cholesterol, myo-Inositol-2-monophosphate, 4-Hydroxysphinganine, Ceramide (d18:1, C24:1), Ceramide (d18:2, C24:0), Sphingomyelin (d18:1, C16:0), 1,2-Dioleoyl-sn-glycero-3-phosphatidyl-L-serine, 18:1 Lyso Phosphatidylcholine, Coenzyme Q9, Glucose, Glycerol, Glycerophosphates, Phosphate, 5-Methoxysphingosine, erythrosphingosine, Threonine, Diacylglycerides and Triacylglycerides.