AURORA Tutorial

Auditable Unified Retrieval and Optimization of Rational Analogs

Before proceeding with the molecule generation and searching, you need to Sign In and obtain your API key.

1. Molecule Generation

Overview:

This tool transforms one molecule (source) into another (target) by generating chemically valid intermediate structures. It identifies a common core between the two molecules and progressively modifies the structure while logging all intermediates.

Requirements:

  • (i)Source SMILES: Your starting molecule
  • (ii)Target SMILES: Your desired end molecule

*Preferably, the molecules with the known and desired activity

Outcome:

A CSV file containing following columns:

  • (i)smiles: SMILES string of the generated molecule
  • (ii)sa_score: Synthetic Accessibility score of the generated molecule
  • (iii)sim_source: Similarity of the generated molecule to the Source
  • (iv)sim_target: Similarity of generated molecule to Target
  • (v)composite_score: Quality score assigned to the particular molecule.

Important Parameters:

(i) Search Parameters

ParameterDescriptionDefault ValueTypical Range
BEAM_SIZEControls search breadth (higher = more exploration but slower)200200 – 5000
EPSILONExploration rate for diversity (higher = more random sampling)0.150.05 – 0.25
SIM_THRESHOLDSimilarity threshold for early stopping (%) i.e similarity of generated molecule to Target9590 – 99
MAX_STEPSMaximum number of search steps10000-

(ii) Scoring Weights (Composite Score = α · Sim_target - β · SA + γ · Sim_source):

SymbolParameterMeaningDefault
αALPHAImportance of reaching target1.0
βBETAPenalty for poor synthetic accessibility0.1
γGAMMAEncouragement to stay near source0.5

Important Considerations:

  • •Valid SMILES: Ensure both input SMILES are chemically valid
  • •Shared Structure: Works best when molecules share a common core
  • •Adjust Weights: Customize α, β, γ based on whether you prioritize reaching target, synthetic feasibility, or staying close to source

2. Database Search

This module automatically searches the generated molecules in the desired databases. We provide you with two choices of the database:

  • (i)ZINC
  • (ii)ChemBL

Input:

The csv file generated via molecule generation pipeline.

Output:

The csv file with few additional columns such as:

  • (i)Hit1(smiles, ZINCID, MW, Tanimoto), Hit2(smiles, ZINCID, MW, Tanimoto) and so on.

Where,

  • •smiles - SMILES of the found hit molecule
  • •ZINCID - ID associated with the found molecule
  • •MW - Molecular weight of the found molecule
  • •Tanimoto - Tanimoto similarity between the found Hit and the query molecule.

User-defined Tanimoto score:

Set this according to your requirements. 1.0 means searching for molecules that are exactly similar.