SIDER

SIDER provides an integrated database of adverse drug reactions (ADRs) for marketed medicines, linking drug–ADR associations, frequency annotations from package inserts, and drug–target relationships to support analysis of ADR mechanisms and prediction of protein–side effect associations.

Key Features:

• Extensive Database: SIDER 4 contains data on 1,430 drugs, 5,880 ADRs, and 140,064 drug–ADR pairs.
• Frequency Data: Side-effect frequency information is extracted from package inserts for 39% of drug–ADR pairs, with 19% of those frequencies reported alongside placebo comparisons.
• Drug Indications (NLP): Natural language processing is used to extract drug indications from package inserts to reduce false positives by distinguishing indications from adverse reactions.
• Integration with Drug Targets: Phenotypic data from clinical sources are combined with known drug–target relationships to identify proteins implicated in causing side effects.
• Predictive Analysis: Analysis of side-effect similarities among drugs is used to predict new drug targets and systematically identify overrepresented protein–side effect combinations, with independent data supporting many links to proteins whose perturbation results in ADRs.
• Experimental Validation: Predictions have been validated experimentally, for example linking serotonin 7 receptor (HTR7) activation to hyperesthesia in mice and showing prevention of the side effect with a selective HTR7 inhibitor.

Scientific Applications:

• Drug Safety Assessment: Provides detailed ADR and frequency data to enhance evaluation of drug safety profiles.
• Mechanistic Studies: Facilitates investigation of molecular mechanisms underlying drug actions and side effects via protein–side effect associations.
• Target Identification: Assists identification of novel drug targets by analyzing side-effect similarity and overrepresented protein–side effect combinations.
• Clinical Trial Design: Informs clinical trial planning by supplying potential ADRs and their reported frequencies, including placebo comparisons where available.

Methodology:

Extraction of side-effect frequency data from package inserts; natural language processing to extract drug indications; integration of clinical phenotypic data with known drug–target relationships; analysis of side-effect similarities to predict targets and identification of overrepresented protein–side effect combinations.