H O W I S I T D O N E ?
1. Provide a reference-molecule and molecule candidates for substition.
2. Select a filter method.
Some filters allow further customization
such as attribute selecting and significance.
Multiple filters can be selected to run in succession.
3. Start filteration.
Results of the filtered molecule candidates will be recieved in Email.
4. Upon recieving the results,
Visual presentation of the molecules is available in our site, simplifying molecule comparisons.
A B O U T O U R M E T H O D S
O U R F I L T E R S
We have 2 main filtration methods consisting of
The NLP model and attribute analysis model
N L P M O D E L
The NLP filter is designed to find molecules with the highest similarity overall amongst all attributes.
The filter was trained on over
600 Million molecules multiple times, discovering many attributes along the way.
This is done by using morgan fingerprints in order to convert molecules into sentences, and using said sentences in order to train a NLP model, specifically, a Word2Vec model.
This model can create representative embedding for each type of structural feature a molecule that it was trained on has, and use that knowledge to compare between different molecules.
A T T R I B U T E A N A L Y S I S M O D E L
The Attribute Analysis filter enables the user to select from over 50 prefedined attributes to be taken into consideration in the filteration.
With this in mind, the user select the attributes that matter to him whilst ignoring those that don't.
Furthermore, the user can select the impact of each of the selected attribute, allowing the user to define the significance of each of the selected attributes.
​
