SimBioSys Inc.



LASSO (Ligand Activity in Surface Similarity Order) is a ligand-based virtual screening tool. It allows users to screen rapidly large libraries of compounds for molecules that share surface properties with a set of known active ligands. LASSO can serve as a pre-screening phase prior to HTS experiments, or prior to more detailed and computationally expensive calculations.

The Algorithm

The LASSO algorithm relies on QSAR descriptors that are calculated for each of the known actives, as well as for the screened compounds. The QSAR descriptor is a vector that counts the ISPs (interaction surface points), or chemical feature flags, on the surface of the molecules. A neural network algorithm then finds common patterns among the active molecules that distinct them from inactives compounds(decoys). The patterns are used to rank each of the screened molecules on a scale between 0 (inactive) and 1 (active).


LASSO's QSAR descriptors are effectively conformation independent - they capture the types of interactions facilitated by the ligand, but do not consider their spatial distribution. The simplistic description removes biases toward molecules that are geometrically similar to the set of actives, and allows scaffold hopping - the identification of potential actives that are structurally distinct from the training set.

LASSO's Main Benefits

  • Ease of use - To run a LASSO screening all you need is a set of active ligands and a library of compounds. The input files should include 3D structures, but the specific poses are not important. The calculation itself is highly automated.
  • Speed - Training a LASSO filter using a set of actives takes typically a few minutes. You can preprocess large libraries of molecules, which will allow you to run the actual screening in seconds or minutes even for very large collections of compounds.
  • Scaffold-hopping - The coarse grained QSAR descriptors facilitate scaffold hopping and good enrichments even for small training sets of actives.
[LASSO Links]