Home
News
Tech Grid
Data & Analytics
Data Processing Data Management Analytics Data Infrastructure Data Integration & ETL Data Governance & Quality Business Intelligence DataOps Data Lakes & Warehouses Data Quality Data Engineering Big Data
Enterprise Tech
Digital Transformation Enterprise Solutions Collaboration & Communication Low-Code/No-Code Automation IT Compliance & Governance Innovation Enterprise AI Data Management HR
Cybersecurity
Risk & Compliance Data Security Identity & Access Management Application Security Threat Detection & Incident Response Threat Intelligence AI Cloud Security Network Security Endpoint Security Edge AI
AI
Ethical AI Agentic AI Enterprise AI AI Assistants Innovation Generative AI Computer Vision Deep Learning Machine Learning Robotics & Automation LLMs Document Intelligence Business Intelligence Low-Code/No-Code Edge AI Automation NLP AI Cloud
Cloud
Cloud AI Cloud Migration Cloud Security Cloud Native Hybrid & Multicloud Cloud Architecture Edge Computing
IT & Networking
IT Automation Network Monitoring & Management IT Support & Service Management IT Infrastructure & Ops IT Compliance & Governance Hardware & Devices Virtualization End-User Computing Storage & Backup
Human Resource Technology Agentic AI Robotics & Automation Innovation Enterprise AI AI Assistants Enterprise Solutions Generative AI Regulatory & Compliance Network Security Collaboration & Communication Business Intelligence Leadership Artificial Intelligence Cloud
Finance
Insurance Investment Banking Financial Services Security Payments & Wallets Decentralized Finance Blockchain
HR
Talent Acquisition Workforce Management AI HCM HR Cloud Learning & Development Payroll & Benefits HR Analytics HR Automation Employee Experience Employee Wellness
Marketing
AI Customer Engagement Advertising Email Marketing CRM Customer Experience Data Management Sales Content Management Marketing Automation Digital Marketing Supply Chain Management Communications Business Intelligence Digital Experience SEO/SEM Digital Transformation Marketing Cloud Content Marketing E-commerce
Consumer Tech
Smart Home Technology Home Appliances Consumer Health AI
Interviews
Think Stack
Press Releases
Articles
Resources
  • Deep Learning

AI in Bioinformatics: Overcoming Pitfalls Webinar


AI in Bioinformatics: Overcoming Pitfalls Webinar
  • Source: Source Logo
  • |
  • June 19, 2025

Xtalks announced a webinar hosted by BullFrog AI to address critical pitfalls in AI-driven bioinformatics workflows. Scheduled for June 23, 2025, the session will explore strategies to maintain statistical integrity, interpret machine learning (ML) outputs accurately, and apply generative AI effectively, ensuring robust outcomes in drug discovery and clinical research.

Quick Intel

  • Xtalks hosts webinar on AI pitfalls in bioinformatics on June 23, 2025.

  • Focuses on compositional data, ML feature importance, and generative AI use.

  • Offers strategies to avoid errors in RNA-seq and other biological data analyses.

  • Introduces rigorous methods like bootstrapped permutation testing.

  • Enhances reproducibility in AI-driven ranking and prioritization tasks.

  • Led by Juan Felipe Beltrán, Director of AI at BullFrog AI.

Addressing AI Challenges in Bioinformatics

The webinar, announced by Xtalks on June 9, 2025, in Toronto, aims to strengthen bioinformatics practices by tackling common analytical pitfalls. As computational biology increasingly relies on statistical methods, ML, and generative AI to process high-throughput datasets, errors in data handling and interpretation can undermine results. This session, led by Juan Felipe Beltrán of BullFrog AI, will provide actionable solutions to enhance trust and reproducibility in drug discovery and clinical development.

Handling Compositional Data Correctly

RNA sequencing and similar datasets produce compositional data, where values represent parts of a whole. A frequent error is filtering out low-count or low-variability genes, which disrupts data integrity and skews results. “While this may seem helpful, it disrupts the balance of the dataset (known as compositional closure) and can lead to misleading statistical results,” notes the webinar overview. The session will demonstrate using a residual category to preserve data balance and applying PFLog1PF transformation to manage zero values, improving accuracy in analyses like PCA and clustering. These methods are vital for RNA-seq, microbiome profiling, and other fields like flow cytometry.

Interpreting ML Feature Importance

ML models often produce feature importance scores to identify key variables, such as genes or clinical markers. However, these scores can vary by algorithm or software and lack built-in uncertainty measures, leading to overinterpretation. The webinar will introduce a statistical framework, including bootstrapped permutation testing, to rigorously evaluate these scores. “By applying methods like bootstrapped permutation testing, researchers can better understand whether the importance of a given feature reflects a real biological signal or is from random noise,” explains the session description. This approach ensures more reliable biological insights.

Optimizing Generative AI for Ranking

Generative AI, including large language models, is increasingly used for ranking biological entities like biomarkers or drug candidates. Unstructured use can yield inconsistent or biased results. The webinar will present a structured approach using pairwise comparisons and the Bradley-Terry model to ensure reproducible rankings with clear confidence measures. This method enhances the reliability of AI-assisted prioritization in bioinformatics workflows, supporting critical decision-making.

Impact on Drug Discovery and Clinical Research

The session emphasizes practical, reproducible strategies to build trust in AI-driven bioinformatics. By addressing pitfalls in compositional data, ML interpretation, and generative AI applications, researchers can improve the accuracy of analytical results. These advancements are crucial for drug discovery and clinical development, where reliable data drives critical decisions and accelerates therapeutic innovation.

This Xtalks webinar offers essential guidance for bioinformatics professionals seeking to navigate AI challenges. By adopting robust analytical practices, researchers can enhance the reliability of their workflows, ultimately advancing the development of life-saving treatments.

 

ABOUT XTALKS

Xtalks, powered by Honeycomb Worldwide Inc., is a leading provider of educational webinars and digital content to the global life science, food, healthcare and medical device communities. Every year, thousands of industry practitioners (from pharmaceutical, biotechnology, food, healthcare and medical device companies, private & academic research institutions, healthcare centers, etc.) turn to Xtalks for access to quality content. Xtalks helps professionals stay current with industry developments, regulations and jobs. Xtalks webinars also provide perspectives on key issues from top industry thought leaders and service providers.

News Disclaimer
  • Share