understanding.bio presents
PhenQSAR Logo
Phenotypic Drug Discovery Platform
DEMO 40 Features Dataset
LAUNCH DEMO

About PhenQSAR

PhenQSAR is a platform that finds potential drug candidates by comparing how compounds affect cells. The basic idea is simple: if a compound changes cells in the same way as turning a gene on or off, that compound might be useful for targeting that gene.

How It Works

The platform compares how compounds change cell appearance against what happens when genes are knocked out (CRISPR) or overexpressed (ORF). This helps find molecules that either copy or oppose the effects of genetic changes.

Key Features

Gene-Based Search

Enter a gene name to find compounds that affect cells in similar or opposite ways to that gene.

Gene Knockout & Overexpression

Compare compounds against both gene knockout and gene overexpression data to find different types of useful compounds.

Chemical Structure Matching

Find compounds that are chemically similar to known drugs by comparing their molecular structures.

Visual Results

View molecular structures, similarity scores, and summary statistics in an easy-to-understand format.

Known Drug Database

Compare results with existing drugs to validate findings and identify proven compounds.

Similarity Analysis

Calculate how similar compounds are to gene effects and assess their drug-like properties.

Dataset Overview

40 Cell Painting Features
116K+ Compounds
11K+ Known Drug Hits
138K+ Profiles (Demo)

Data Processing

The dataset is processed through multiple steps:

  • wellpos - Position correction
  • var - Variance adjustment
  • mad - Outlier detection
  • outlier - Outlier removal
  • featselect - Feature selection
  • sphering - Data normalization
  • harmony - Batch correction

Technical Implementation

Similarity Scoring

Cell Similarity: Compares how similar cell changes are between genes and compounds. Scores range from -1 to +1, where positive scores mean similar effects and negative scores mean opposite effects.

Chemical Similarity: Compares molecular structures to find chemically similar compounds. A score above 0.55 indicates high structural similarity.

Analysis Settings

  • Minimum similarity: Greater than 0.2
  • Gene knockout results: Top and bottom 100 compounds
  • Gene overexpression results: Top and bottom 20 compounds
  • Compound size filter: Molecular weight less than 1000
  • High similarity threshold: Greater than 0.5