Summary
- Used ANNE technology to develop regression and classification models
- Curation identified activity cliffs and questionable LD50 values
- Model applicability domain is defined by the minimum and maximum descriptor values in the training set
- Confidence estimates are included in binary classification models
- Submitted two models for each of the five endpoints
NIH: Predictive Models for Acute Oral Systemic Toxicity Workshop, April 11-12, 2018, Bethesda, MD
By Michael Lawless, Pankaj R. Daga, Marvin Waldman, Robert Fraczkiewicz, Robert D. Clark, John DiBella, and Michael B. Bolger