Reviewed by Chris Earnshaw
CGE Computational Chemistry, UK
Structures and data are easily imported and StarDrop can rapidly calculate a number of properties together with uncertainty values reflecting the confidence of the prediction. The built in auto-modeller feature makes it easy to generate local models based on user-supplied data. Interactive plotting tools (histograms, X-Y plots, etc) are provided to visualise data in the table; selecting a molecule highlights the corresponding points in the plot and vice versa. A 2D representation of chemical space can be generated to visualise clusters and the position of specific molecules within them.
Another clever feature is the 'glowing molecule' which highlights the influence of each atom on the modelled properties, allowing medicinal chemists to interactively explore the effect of structural changes with immediate visual feedback. The remaining aspect is the ability to define a target profile based on the data in the table and apply 'probabilistic scoring' (including the uncertainty values) to identify the potentially interesting compounds. Once the target profile has been defined, a column is added to the table giving the score for each compound and a histogram showing the individual profile. The data cells in the table are colour coded to highlight contributions to the profile. Conventionally the highest scoring compounds would be taken forward; StarDrop makes it easy to visualise the scores using a 'snake' plot with error bars which indicate the uncertainty. This can identify compounds that still have a good chance of being successful but which would perhaps have been rejected on the scores alone.
StarDrop is a powerful but user-friendly software package providing useful models and analysis tools; I would strongly recommend it to medicinal chemists and other drug discovery scientists.
For more information contact firstname.lastname@example.org or visit the website.
Software for all stages of drug discrovery
External links will open in a new browser window