Nat. be used to generate drug repurposing hypotheses, using Alzheimers disease as a use-case. Availability: https://www.ebi.ac.uk/chembl/ftc; https://github.com/loopasam/ftc. Contact: ku.ca.ibe@tesorc Supplementary information: Supplementary data are available at online. 1 MOTIVATION Drug repurposing is the use of known active compounds for new therapeutic indications (Sanseau and Koehler, 2011). When administered in a living organism, a compound can indeed play various functions and affect different biological processes [called mode of action (MoA)]; accurately identifying these different functions helps to predict the potential side-effects a drug could have and can also lead to interesting repurposing opportunities (Medina-Franco (2011) or Andronis (2011) for recent reviews]. Most methods operate on the profiles of physicochemical descriptors derived from molecular structures (Haupt and Schroeder, 2011). Other methods characterize Pexidartinib (PLX3397) the drugs on more abstract levels, such as the gene expression signature (Iorio (mouse model) the potential of the drug and other histone deacetylase inhibitors in regards to memory deficit (Kilgore (2012). The FTC mostly differentiates itself from these projects by providing a whole set of new categories on the top of the integrated information, dedicated to tackle a very specific problem: drug repositioning. 3.1 Biological assumptions An asset of the FTC is usually its ability to handle efficiently categorical data: classes and relationships are accurately defined, in order to classify compounds based on the semantics of their relations. The properties linking drugs to their respective protein targets (positive and negative perturbations) are, however, simplistic. At the time being, no consideration is usually given regarding the binding strength between the drug and the proteins, yet it is a key factor to derive potent and specific activities in the human body. This is also the case for other types of numerical data, such as the dosage; the FTC can predict a role for a drug, yet it cannot provide any information about the concentration Pexidartinib (PLX3397) or the administration route necessary to obtain the potential effects. The current relations between targets and their involvement in biological processes are also not a fully accurate representation of the Pexidartinib (PLX3397) biological phenomenon. In a cell, specific domains of the protein could mediate different functions. Only one of such activity Pexidartinib (PLX3397) types can sometimes be inhibited by a drug (Kruger em et al. /em , 2012), yet we are assuming in the FTC that as long as a drug affects a protein, it can therefore alter all its known functions. These limitations come from the semantics behind the axioms structuring the classification themselves based on the information available from the databases. Despite entailing not entirely accurately the biochemical reality, the axioms help to generate a larger number of hypotheses, the primary goal of the FTC. The dosage issue is partially addressed by the regulator pattern (see Section 3.1 of Supplementary Material): it should be easier to experimentally adjust the concentration of the compounds classified as pro- or anti- biological process agents in order to modulate a physiological effect. The predictions generated by the FTC depend around the resolution of the curated information released by the original data providers. Erroneous or missing information will lead to misclassification by the reasoner. Some expected outcomes are also missing from the predictions; sildenafil for instance was expected to be classified as pro-penile erection agent (FTC_A0043084), yet the lack of appropriate GO annotation prevents it. After discussion with the GOA curation team, a manual annotation can only be asserted based on published experimental results. No document was found to support the involvement of the cGMP-specific 3,5-cyclic.305C320. available at online. 1 MOTIVATION Drug repurposing is the use of known active compounds for new therapeutic indications (Sanseau and Koehler, 2011). When administered in a living organism, a compound can indeed play various functions and affect different biological processes [called mode of action (MoA)]; accurately identifying these different functions helps to predict the potential side-effects a drug could have and can also lead to interesting repurposing opportunities (Medina-Franco (2011) or Andronis (2011) for recent reviews]. Most methods operate on the profiles of physicochemical descriptors derived from molecular structures (Haupt and Schroeder, 2011). Other methods characterize the drugs on more abstract levels, such as the gene expression signature (Iorio (mouse model) the potential of the drug and other histone deacetylase inhibitors in regards to memory deficit (Kilgore (2012). The FTC mostly differentiates itself from these projects by providing a whole set of new categories on the top of the integrated information, dedicated to tackle a very specific problem: drug repositioning. 3.1 Biological assumptions An asset of the FTC is usually its ability to handle efficiently categorical data: classes and relationships are accurately defined, in order to classify compounds based on the semantics of their relations. The properties linking drugs to their respective protein targets (positive and negative perturbations) are, however, simplistic. At the time being, no consideration is usually given regarding the binding strength between the drug and the proteins, yet it is a key factor to derive potent and specific activities in the human body. This is also the case for other types of numerical data, such as the dosage; the FTC can predict a role for a drug, yet it cannot provide any information about the concentration or the administration route necessary to obtain the potential effects. The current relations between targets and their involvement in biological processes are also not a fully accurate representation of the biological phenomenon. In a cell, specific domains of the protein could mediate different functions. Only one of such activity types can sometimes be inhibited by a drug (Kruger em et al. /em , 2012), yet we are assuming in the FTC that as long as a drug affects a protein, it can therefore alter all its known functions. These limitations come from the semantics behind the axioms structuring the classification themselves based on the information available from the databases. Despite entailing not entirely accurately the biochemical reality, the axioms help to generate a larger number of hypotheses, the primary goal of the FTC. The dosage issue is partially addressed by the regulator pattern (see Section 3.1 of Supplementary Material): it should be easier to experimentally adjust the concentration of the compounds classified as pro- or anti- biological process agents in order to modulate a physiological effect. The predictions generated by the FTC depend around the resolution of the curated information released by the original data providers. Erroneous or missing information will lead to misclassification by the reasoner. Some anticipated outcomes will also be missing through the predictions; sildenafil for example was likely to become categorized as pro-penile erection agent (FTC_A0043084), the lack of suitable Move annotation prevents it. After dialogue using the GOA curation group, a manual annotation can only just become asserted predicated on released experimental outcomes. No record was found to aid the involvement from the cGMP-specific 3,5-cyclic phosphodiesterase (sildenafils primary focus on) in the adverse TLN1 rules of penile erection (Move:0060407), zero annotation could be produced therefore. Further work could possibly be completed in this path, by looking to infer even more annotations or utilizing the electronically produced types instantly, to be able to generate broader however much less plausible repurposing hypotheses potentially. 3.2 Interpreting the evaluation From the evaluation, the high recall worth (89%) supports the theory behind the automated build from the FTC: the info from different repositories funded and curated in parallel, could be integrated to make a fresh source automatically. This fresh classification (FTC) consists of a lot of the known info present in.