In the approach described here two types of terms (bioentities) have been used: those referred to chemical products and those related to diseases. The terms were extracted from PubMed abstracts, and are related to human genes by a score derived from the frequency of gene-term co-occurrences and depending on their proximity within the text. The scores are derived from a Z-statistic that estimates how unlikely it is to observe a certain level of co-occurrences to happen by chance (Andrade and Valencia, 1998). The gene-bioentity correspondence tables with the respective scores were obtained using the AKS software (available at: http://www.bioalma.com/aks2/) and are freely available in the GeneCards (Safran et al., 2003) database (http://www.genecards.org/). Contrarily to the case of GO and other similar functional categories, bioentities are not discrete classes. The membership of a gene to a given bioentity is conditioned through the scores.