Case Study 1: Ebola drug discovery

 
Cells were treated and then challenged with Ebola virus encoding GFP. Infection efficiency was calculated as infected cells (expressing GFP)/total cells and normalized to infection efficiency seen in the untreated control. Shown is one representative experiment where each point is the average of 3 independent measurements of infection +/- standard deviation. Dose response curves were fitted by non-linear regression.

Cells were treated and then challenged with Ebola virus encoding GFP. Infection efficiency was calculated as infected cells (expressing GFP)/total cells and normalized to infection efficiency seen in the untreated control. Shown is one representative experiment where each point is the average of 3 independent measurements of infection +/- standard deviation. Dose response curves were fitted by non-linear regression.

Machine Learning Model to In Vitro Actives

An early success validating our approach was demonstrated for Ebola virus (EBOV). We used public data to generate Bayesian models for the viral pseudotype entry assay and the EBOV replication assay. These models were used to score the MicroSource Spectrum set of 2320 molecules. We selected 3 compounds for testing in vitro. All three were returned as actives.

Virus challenge was on day 0, and once-daily i.p. dosing of tilorone at 30mg/kg was given at 2 h postinfection (h p.i.) (red), 30 mg/kg at 24 h p.i. (green), 60 mg/kg at 2 h p.i. (purple), 60 mg/kg at 24 h p.i. (orange), and 60 mg/kg at 48 h p.i. (black), or vehicle was given (blue). (A) Kaplan-Meier survival curves. *, P < 0.01 for treatment versus vehicle groups, assessed by log-rank analysis using the Dunnett-Hsu procedure to adjust for multiple comparisons.

Virus challenge was on day 0, and once-daily i.p. dosing of tilorone at 30mg/kg was given at 2 h postinfection (h p.i.) (red), 30 mg/kg at 24 h p.i. (green), 60 mg/kg at 2 h p.i. (purple), 60 mg/kg at 24 h p.i. (orange), and 60 mg/kg at 48 h p.i. (black), or vehicle was given (blue). (A) Kaplan-Meier survival curves. *, P < 0.01 for treatment versus vehicle groups, assessed by log-rank analysis using the Dunnett-Hsu procedure to adjust for multiple comparisons.