Automated Literature Reviews in Pharma Using AI Agents

Introduction

This workflow outlines the steps for conducting automated literature reviews and research synthesis in the pharmaceutical industry, emphasizing the integration of AI agents to enhance efficiency and accuracy throughout the process.

1. Define Research Question and Scope

• Human researchers establish the research question, scope, and inclusion/exclusion criteria.
• AI Agent Assistance: An AI planning agent can refine the research question by analyzing recent trends and suggesting relevant sub-topics or perspectives.

2. Search and Retrieve Literature

• Conduct comprehensive searches across multiple databases.
• AI Agent Integration:
  • PICO Portal: An evidence synthesis platform utilizing machine learning to identify relevant citations.
  • DistillerSR: Automates literature collection and screening using AI and intelligent workflows.

3. Screen and Select Relevant Studies

• Review titles and abstracts to determine relevance.
• AI Agent Integration:
  • LiteRev: Employs natural language processing and machine learning to streamline screening, suggesting relevant papers with a recall rate of 87.5% compared to manual screening.
  • DistillerSR’s AI-Powered Screening: Can reduce the screening burden by up to 60%.

4. Extract Data

• Extract relevant data from selected full-text articles.
• AI Agent Integration:
  • JBI Sumari: Facilitates the entire systematic review process, including data extraction.
  • InsightRX Apollo-AI: Enhances analytical capabilities for pharmacokinetic and pharmacodynamic analyses.

5. Assess Study Quality

• Evaluate methodological quality and risk of bias.
• AI Agent Integration: An AI agent can be developed to assess study quality based on predefined criteria, flagging potential issues for human review.

6. Synthesize Evidence

• Analyze and synthesize extracted data.
• AI Agent Integration:
  • Topic modeling algorithms like latent Dirichlet allocation (LDA) can identify key research themes and emerging trends.
  • Association rule mining techniques can discover relationships between studies or data points.

7. Interpret Results and Draw Conclusions

• Interpret synthesized evidence and draw conclusions.
• AI Agent Integration: An AI analysis agent can provide initial interpretations and suggest potential conclusions based on the synthesized data, which human researchers can then review and refine.

8. Write and Publish Review

• Prepare the final manuscript for publication.
• AI Agent Integration:
  • AI writing assistants can help draft sections of the manuscript.
  • Automated citation management tools ensure proper formatting and completeness of references.

9. Update and Maintain Review

• Continuously update the review as new evidence emerges.
• AI Agent Integration:
  • DistillerSR’s LitConnect: Automatically imports newly published references to keep literature reviews up-to-date.
  • An AI monitoring agent can scan for new publications and alert researchers when significant new evidence is available.

Process Improvements with AI Agents

1. Enhanced Speed and Efficiency: AI agents can significantly accelerate the review process. For example, LiteRev demonstrated a work saved over sampling of 56% compared to manual methods.
2. Improved Accuracy: AI-powered tools like DistillerSR’s duplicate detection can prevent skew and bias caused by including studies more than once.
3. Scalability: AI agents can handle large volumes of data more efficiently than human reviewers, allowing for more comprehensive reviews.
4. Consistency: AI agents can apply inclusion/exclusion criteria consistently across all studies, reducing human bias.
5. Real-time Updates: Tools like DistillerSR’s LitConnect enable living reviews that stay current with the latest research.
6. Enhanced Analysis: Advanced AI techniques like topic modeling and association rule mining can uncover patterns and relationships that might be missed in manual reviews.
7. Customization: Platforms like Allex.ai can be tailored to specific pharmaceutical research needs, integrating with existing workflows.
8. Collaboration Support: Many AI-powered tools facilitate team collaboration, allowing multiple researchers to work efficiently on the same review.

By integrating these AI agents and tools, pharmaceutical companies can significantly streamline their literature review and research synthesis processes, leading to faster insights, more comprehensive analyses, and ultimately, accelerated drug discovery and development.