BASF Metabolome Solutions GmbH
Pattern Ranking
The idea behind the pattern ranking is the assumption that compounds / chemicals that produce a specific form of toxicity also share a common set of endogenous metabolites that is changed as a consequence of the administration of the chemical. This common set of metabolites is referred to as a “pattern”.
Principle of Pattern Ranking:
The comparison between a test substance and the pattern of metabolite changes predictive for a particular MOA is defined as “pattern ranking”. The result of this pattern ranking process is a table displaying the predefined patterns in the first column, followed by metadata information, the number of compounds in the pattern, quantiles belonging to it and parameters such as the median r(i) of the N correlations, the correction factor for r(i) (r(i)CF) and the corrected median r(i)corr. Further, a plot describes the variability of the metabolites of the reference compounds used to create the pattern as a black box, overlaid by the distribution of the metabolites of the test compound as a red dot including its standard deviation. Moreover, a red triangle indicates the position of anchor metabolites which have to be changed in the right direct in order for a pattern to be considered as a match.
Figure 1: Pattern ranking
Correlation of a treatment against all final patterns in the database and ranking according to correlation score.
Generation of Patterns:
For the generation of a new pattern, mostly, at least 4 elements are needed: Common metabolites from at least 3 different chemicals that share the same toxicological mode of action (reference compounds) are determined (sex-separated and for at least 2 time-points unless toxicity indicates that the MOA can only be expected at 1 time point). Afterwards, a first validation step is performed by presenting the pattern to an expert metabolomics panel. Then, the pattern is validated against the MetaMap® Tox database, in which at least one additional compound (at 90% fraction of metabolites) has to be identified (sensitivity check). Compounds not sharing the specific mode of action, however, should not be identified (selectivity check) (Bennard van Ravenzwaay et al. 2015). If possible, the pattern is confirmed biochemically by including anchor metabolites or if applicable, an antagonist to the mode of action is assessed which should have its metabolites regulated in the opposite direction.
Figure 2: The process of pattern generation
Analysis of Patterns:
For a pattern ranking analysis, a comparison of the metabolite changes of a test substance with all predefined final patterns for metabolite changes which are predictive for particular effects, is conducted. Based on the resulting ranked list of patterns, each pattern is analyzed individually in MetaMap® Tox. The r(i) CF denotes the amount of metabolites of the respective pattern that are significantly changed. To analyze the pattern, each pattern is analyzed and the number of metabolites regulated in the right direction are counted. If compared to the pattern, the number of significantly changed metabolites in the right direction are 90% or more, the pattern is considered a match. Subsequently, weak matches are defined by 75% - 90% metabolites significantly changed in the right direction. >50% - 75% of metabolites are considered as equivocal and less than 50% of metabolites significantly changed in the right direction as a mismatch. Furthermore, for a pattern to be classified as a match, an anchor metabolite - if present - has to be compulsorily changed in the right direction, otherwise the pattern is instantly classified as a mismatch. To judge a test compound, the 20 top-ranking patterns are assessed at least.