Ontomine – in silico prediction of toxicity, adverse effects and biological activities

Introdction to Ontomine

Computational approaches for bioactivity, drug target and toxicity spectrum predictions has value in various areas like drug target identification, lead discovery and identification, buying chemical libraries, exploring in-house corporate molecular libraries, compound selection for vHTS or HTS, estimating the specificity and target ranges for developed compounds and even hinting for possibilities for drug repositioning.

Ontomine is a software used for drug target, bio-activity, toxicity and adverse side effect prediction. We have collected and hand curated an extensive knowledgebase of experimentally detected biological activities for more than 100.000 small molecules. The used training set molecules are diverse in chemical structures, functional group features and molecular weight. The knowledgebase contains a large number of biological activity endpoints and small molecules, that have been observed to have an effect on the given activity, which include binding to drug targets, physiological effects of drugs, therapeutic uses, observed toxicities and adverse side effects. Altogether, there are almost 3000 endpoint prediction classes in the Ontomine knowledgebase.

Based on this collected data, molecular fingerprints are calculated for all these endpoints. The fitting algorithm enables the prediction of bio-activities for novel molecules and reporting the molecular basis of the predictions. The OntoMine prediction algorithm is based on the correlation between the generated fingerprint of each prediction endpoint and the calculated properties for the analyzed new small molecule. Nominal prediction confidence levels are calculated on the basis of similarity threshold between the molecule and rules applied from given prediction endpoint. Results are ranked on the basis of the confidence level for a given class.

Ontomine predicted Drug Targets

There are about 1700 drug target endpoints in the Ontomine knowledgebase. Some of them are listed below, please contact us for the current complete list.

Enzymes: Major families including Proteases, Kinases, Oxidoreductases, Cholinesterases, Decarboxylases, Lyases, Dehydrogenases, Helicases, Hydrolases, Isomerases, Ligases, Transferases, Phosphodiestreases, Reductases, Synthases
Ion channels including Potassium channels, Sodium channels, Chloride channels
Receptors: Major receptor classes including GPCRs, Kinase receptors, Fibroblast growth factor receptors
Receptor subclasses including Alpha-adrenergic, Dopamine, Adenosine, Histamine, Glutamate, Calcitonin, Cannabinoid, Endothelin, Metabotropic, Muscarinic Acetylcholine, Somatostatin
Drug targets: some examples of drug targets include STAT 1 & inhibitors, STAT3 activator, tissue-type and urokinase-type plasminogen activator inhibitors, acetylcholinesterase inhibitors, Histone deacetylase 1,2,3,6,8,9 inhibitors, estradiol 17-beta dehydrogenase 1 inhibitors, macrophage colony stimulating factor 1 receptor inhibitors, monoamine oxidase inhibitors, myosin IIIA and IIIB inhibitors, Pim-2 and -3 oncogene inhibitors, Vascular endothelial growth factor receptors 1,2 and 3 inhibitors.
The patterns for Agonists/Antagonists and Activators/ Inhibitors for each of the above are available.

Ontomine case study Ontomine was used to predict the target classes for three marketed. In the chart are relevant Ontomine toxicity predictions for these drugs.

Ontomine predicted pharmacologic actions

There are over 300 pharmacologic action endpoints in the OntoMine knowledgebase. Some of them are listed below, please contact us for the current complete list.

Molecular Mechanisms of Pharmacological Action, like Anti-Oxidants, Enzyme Reactivators, Physiological Effects of Drugs, like Hypoglycemic Agents and Peripheral Nervous System Agents
Therapeutic Uses, like Anti-Allergic Agents, Anti-Infective Agents, Anti-Inflammatory Agents
Other examples of pharmacologic actions include bone density conservation agent, androgen antagonists, estrogen receptor modulators, prostaglandin antagonists, interferon inducers, neuromuscular blocking agents, neuromuscular depolarizing agents, anti-cholesteremic agents, anti-angiogenesis agent, anti-neoplastic agents, anti-neovascularisation agents, anti-Parkinson agents, anti-anxiety agents, skeletal muscle relaxants.

Ontomine case study 100 compounds, not present in Ontomine knowlegde base and with experimentally verified target protein classes were used to test Ontomine’s target predictions. The only input to Ontomine was the small molecule compounds structures. The chart indicates the accuracy whether Ontomine was able to predict the targets for these 5 major target classes/proteins. Predictions include HIGH, MEDIUM and LOW nominal classes.

Ontomine predicted toxicities and adverse effects

There are over 100 toxicities and adverse effects endpoints in the OntoMine knowledgebase. Some of them are listed below, please contact us for the current complete list.

Acute Toxicity
Carcinogenicity
Cytotoxicity
Eye Irritancy
Immunotoxicity: Immuno Supressant, AutoImmune Diseases
Mutagenesis and Genetic Toxicity: Genotoxicity, Mutagenicity
Reproductive Toxicity: Developmental Toxicity, Female Reproductive Toxins, Male Reproductive Toxins, Teratogenicity, Embryo Toxicity
Respiratory Toxins: Pulmonary Toxicity, Nasal Irritancy
Side Effects: Dizziness, Bloating, Abdominal Pain, Hypertension, Diarrhea
Specific Target Organ Toxicity: Cardio Toxins, Circulatory Toxins, CNS Toxins. Digestive Toxins, Endocrine Toxins, Excretory Toxins, Hepatotoxicity, Integumentary Toxins, Sensory Toxins

Ontomine case study Early 2011 four drugs were banned from the market (Nimesulide, Cisapride, Phenylpropanolamine & Sibutramine) due to their high toxic effects (carcinogenic, hepatotoxicity, heart toxin in case of Sibutramine etc). In the chart are relevant Ontomine toxicity predictions for these drugs.

Ontomine and comparison to other existing methods

Other currently used bioactivity prediction approaches can be classified into three categories:

graph theoretic and substructure, topological and physicochemical chemical descriptors
maximum common substructures, fingerprints similarities and machine learning methods
atom neighborhood methods

However, these methods are currently limited by factors like distance metrics dependance, cut-offs not being representative of activity, hard to build QSAR models at several different levels, e.g. drug targets, bio processes and therapeutic area as well as interpretation of results.

Ontomine is alternative to these methods, transforming the structural information for chemically, biologically or pharmacologically related molecules to a hierarchical schema of concepts and descriptors. Ontomine discovers patterns in the related schema and predicts biological activity, using rules inferred from analyzing the patterns.

Ontomine case study 100 compounds, not present in Ontomine knowlegde base and with experimentally observed toxicities were used to test Ontomine’s toxicity predictions. The only input to Ontomine was the small molecule compounds structures. The chart indicates the accuracy whether Ontomine was able to predict the observed toxicities for the 5 major toxicity classes. Predictions include HIGH, MEDIUM and LOW nominal classes.

Ontomine patterns represent “necessary but not sufficient” conditions for bioactivity irrespective of a scaffold. Explicit scaffold hopping can be performed by generating constitutional isomers and selecting interesting molecules using the rule base. The results of Ontomine are presented as simple understandable patterns representative of activity. Further, Ontomine predictions are scalable to millions of molecules. In addition, the curated Ontomine datasets can be appended with internal datasets of measured molecule activities of in-house compound collections.

Ontomine application areas

In silico screening workflows can be composed of Ontomine-ADME for physicochemical and ADME properties, Ontomine for bioactivity, adverse effect and toxicity predictions as well as other common (like Lipinski’s rules-of-five) and custom filters. Follow this link for more on our QSAR ADME property prediction approches.

Availability of Ontomine

We offer Ontomine predictions both as a service and as a licensed software. We also offer evaluation of the predictions of Ontomine and Ontomine-ADME software for your own molecules, please contact us for details.

Please follow these link to retrieve a fact sheet or a sample report of our Ontomine software or to request a quote