Pharmaceutical Industry
Permion solves the problem of accurately predicting drug-induced injury in the drug development process.
Applications
Permion Graph ML
Traditional methods for molecular toxicity prediction may not
capture complex interactions between drugs, leading to
inaccurate predictions and potential harm to patients.
Additionally, current processes do not extract enough
information from graph datasets during training of graph neural
networks.
Permion Graph-ML uses graph convolutional neural networks to
model the complex interactions between drugs, resulting in more
accurate toxicity predictions. Permion Graph-MLâs accuracy leads
to a more efficient and effective approach to toxicity
prediction, has the potential to accelerate drug development,
reduce adverse effects, and ultimately improve patient outcomes.
Problem that Permion solves
Current processes do not extract enough information from graph datasets during training of GNNs. With Permion Graph ML, users extract more information leading to: Higher accuracy when predicting properties of graphs. For the molecule-classification task, we reduce prediction error in up to 50% (depending on the dataset) More efficient training, i.e., collect less data to achieve the same level of accuracy.
User Interface
Test drive Permion Graph ML
To demonstrate, we invite pharmaceutical engineers and scientists to see the possibilities how Permion would improve their current processes.
Open Interface View TutorialEfficiency & Effectiveness
The Permion Graph-ML offering has the potential to translate into significant savings for pharmaceutical companies in several ways:
- Faster drug development timelines and lower overall costs
- Higher accuracy, specificity, and sensitivity than traditional methods, potentially reducing the risk of costly adverse events and recalls
- Requires minimal data preparation, reducing the need for expensive data curation resources
Groundbreaking Results: Toxicology Prediction
10 March 2023 - Predicting properties of new and existing chemical compounds is slow and expensive. Existing methods â painstaking lab tests â cannot keep up with the thousands of chemicals that need to be analyzed for drug properties, toxicity, and other interactions. The promise of powerful machine learning has not yet proven up to this challenge â until today.
Permion Graph ML is our fully automated graph machine learning system, its design inspired by the challenge of these and other benchmark graph-format datasets. In chemical compound prediction tasks, Permion Molecular Properties Prediction has beaten the performance of 95% of state-of-the-art methods, in some cases needing 50% less training data while outperforming all published results.