nsides

is a Data Source.

nSIDES is a comprehensive collection of drug side effect and drug interaction resources developed by the Tatonetti Lab. It includes OnSIDES (adverse events from drug labels), KidSIDES (pediatric drug safety signals), AwareDX (sex-specific adverse drug effects), OffSIDES (off-label side effects), TwoSIDES (drug-drug interactions), and ManySIDES (combinations of 3+ drugs).

Domains

pharmacology, drug discovery, health, precision medicine

License

Warning: No license entered

Homepage

nsides

Repository

GitHub

Infores ID

infores:nsides

FAIRsharing ID

Unknown

Product Summary

Products

From this Resource
ID Name URL Category Format Description
nsides.onsides OnSIDES releases Product csv Adverse drug events extracted from FD...
nsides.kidsides KidSIDES ade_nichd.csv.gz (107 B) Product csv Pediatric drug safety signals across ...
nsides.pdsportal PDSPortal pdsportal GraphicalInterface http Interactive RShiny web portal for bro...
nsides.awaredx AwareDX ?p=nsides Product http Drug safety signals with differential...
nsides.onsides.code OnSIDES Code onsides ProcessProduct ❔ Code repository for OnSIDES model tra...
nsides.kidsides.code KidSIDES Code pediatric_ade_database_study ProcessProduct ❔ Code repository for KidSIDES pediatri...

Details

nSIDES

Overview

nSIDES is the home for comprehensive drug side effect and drug interaction resources developed by the Tatonetti Lab at Columbia University. The platform provides multiple interconnected databases and tools for analyzing adverse drug events across different populations and contexts. nSIDES represents one of the most extensive collections of drug safety information available, combining machine learning extraction from drug labels with pharmacovigilance signal detection from post-marketing surveillance data.

The nSIDES ecosystem includes resources for general adverse drug events (OnSIDES), pediatric populations (KidSIDES), sex-specific risks (AwareDX), off-label effects (OffSIDES), drug-drug interactions (TwoSIDES), and higher-order drug combinations (ManySIDES).

Data Content

OnSIDES - Drug Label Adverse Events

OnSIDES extracts adverse drug events from FDA-approved drug labels using a fine-tuned PubMedBERT language model:

  • Coverage: 3.6 million+ drug-ADE pairs
  • Drug ingredients: 2,793 unique compounds
  • Labels processed: 46,686 drug labels from DailyMed
  • Model performance: F1 score of 0.90, AUROC of 0.92 for adverse reactions sections
  • Updates: Quarterly releases
  • International variants: OnSIDES-INTL (Japan, UK, EU labels), OnSIDES-PED (pediatric-specific)
  • Data source: DailyMed structured product labels (November 2023)

KidSIDES - Pediatric Drug Safety

Pediatric drug safety signals across developmental phases:

  • Scope: Drug safety signals specific to childhood developmental stages
  • Method: Pharmacovigilance algorithm applied to post-marketing reports
  • Application: Identifies age-specific adverse event risks
  • Interactive access: PDSPortal RShiny web application
  • Data formats: MySQL, SQLite, and CSV files
  • Coverage: Adverse events analyzed across NICHD developmental phases

AwareDX - Sex-Specific Drug Risks

Machine learning-identified adverse drug effects with differential risk by biological sex:

  • Coverage: 20,817 adverse drug effects with sex-specific risks
  • Focus: Effects disproportionately affecting women
  • Method: ML algorithm accounting for confounding biases
  • Validation: Validated against pharmacogenetic mechanisms of sex-differentially expressed genes
  • Clinical relevance: Women experience 2x risk of ADRs compared to men

OffSIDES and TwoSIDES

Classic resources for off-label side effects and drug-drug interactions:

  • OffSIDES: Off-label drug side effects not listed on FDA labels
  • TwoSIDES: Comprehensive drug-drug-effect relationships
  • Drugs covered: 3,300+ drugs
  • Drug combinations: 63,000+ pairs
  • Adverse reactions: Millions of potential effects
  • Note: Currently being updated for 2022+ with quarterly updates planned

ManySIDES

Side effect signals for combinations of 3 or more drugs:

  • Status: Active development (v0.1 released)
  • Scope: Higher-order drug combination effects
  • Use case: Polypharmacy risk assessment

Key Features

  • Machine learning extraction: Fine-tuned PubMedBERT for label processing
  • Population-specific: Dedicated resources for pediatrics, sex differences
  • Comprehensive coverage: Labels, post-marketing surveillance, interactions
  • Regular updates: Quarterly releases for OnSIDES and derivatives
  • Multiple formats: CSV, SQL, SQLite for flexible integration
  • Interactive tools: Web portals for browsing (PDSPortal)
  • Open source: Code and data freely available
  • Validated methods: Published algorithms with performance metrics

Access Methods

  1. Direct Download: CSV, SQL, and SQLite files from S3 buckets and GitHub
  2. GitHub Releases: Versioned releases with release notes
  3. Interactive Web Portals: PDSPortal for browsing pediatric signals
  4. API Access: (Under development for some components)
  5. Code Repositories: Full source code for reproduction and extension

Technology Stack

  • Machine Learning: Fine-tuned PubMedBERT language model
  • NLP: Named entity recognition and relation extraction
  • Statistics: Propensity score methods, disproportionality analysis
  • Web Applications: RShiny for interactive visualization
  • Data Processing: Python, SQL
  • Model Training: PyTorch/Transformers
  • Data Sources: FDA FAERS, DailyMed, clinical trial databases

Use Cases

  1. Drug Development: Early identification of potential safety signals
  2. Pharmacovigilance: Post-marketing surveillance and signal detection
  3. Precision Medicine: Tailoring drug prescriptions based on sex, age
  4. Clinical Decision Support: Informing prescribing decisions for vulnerable populations
  5. Research: Studying mechanisms of adverse drug events
  6. Regulatory Science: Supporting drug safety assessments
  7. Polypharmacy Analysis: Understanding risks of multi-drug regimens

Model Performance

OnSIDES Extraction Accuracy

  • F1 Score: 0.90 (Adverse Reactions section)
  • AUROC: 0.92 (Adverse Reactions)
  • AUPR: 0.95 (Adverse Reactions)
  • TAC 2017 Evaluation: Micro-F1 0.87, Macro-F1 0.85
  • Boxed Warnings: F1 0.71, AUROC 0.85

Management

Developed and maintained by the Tatonetti Lab at Columbia University, led by Dr. Nicholas P. Tatonetti. The lab specializes in computational pharmacology, pharmacovigilance, and precision medicine applications of machine learning and data science.

  • FAERS: FDA Adverse Event Reporting System, source data for many nSIDES analyses
  • DailyMed: Source of drug labels for OnSIDES extraction
  • SIDER: Alternative drug side effect database
  • DrugBank: Comprehensive drug information database

Last Update

OnSIDES: November 2023 (labels), with quarterly updates planned KidSIDES: Version 0.3 (2019 Q2 data), November 2021 AwareDX: 2020 publication OffSIDES/TwoSIDES: 2022 update in progress

Is this information incorrect or incomplete? Request an update.

Created: October 30, 2025 | Last modified: October 31, 2025