Dr.Sim

Dr.Sim learns similarity measurements between high-throughput transcriptional profiles to improve identification of drug mechanisms of action and therapeutic indications in phenotypic drug discovery.

Key Features:

• Automated Similarity Inference: Employs a learning-based approach to automatically infer similarity measurements between transcriptional profiles, replacing manually defined unsupervised measures.
• Generalized Performance: Demonstrates robust performance across diverse datasets, including publicly available in vitro and in vivo transcriptional perturbation data.
• Facilitation of Phenotypic Drug Discovery: Learns transcriptional similarities to facilitate elucidation of drug mechanisms of action and identification of new drug indications.
• Conceptual Improvement Over Existing Methods: Outperforms traditional unsupervised similarity measures by providing more reliable similarity characterizations for high-dimensional, noisy transcriptional data.

Scientific Applications:

• Connectivity Map and LINCS integration: Applicable to compound perturbation resources such as Connectivity Map (CMap) and Library of Integrated Network-Based Cellular Signatures (LINCS) for mining perturbation datasets.
• Drug annotation and repositioning: Supports drug annotation and repositioning analyses by identifying similar transcriptional responses across compounds and conditions.
• Mechanism of action inference: Aids inference of mechanisms of action by matching transcriptional signatures between perturbations and known agents.
• High-throughput perturbation mining: Enhances computational mining of high-throughput transcriptional perturbation data for phenotypic screening.

Methodology:

Uses a learning-based framework to automatically infer similarity measurements between transcriptional profiles and was evaluated on publicly available in vitro and in vivo datasets, showing improved performance relative to traditional unsupervised measures.