PythRaSh's AI Newsletter

Week of September 23, 2025

Hi, There! This week marks a revolutionary moment in the convergence of artificial intelligence and biological sciences. We're witnessing an unprecedented transformation where AI is no longer just analyzing biological data—it's actively predicting cellular therapies, designing new drugs for human trials, and even creating autonomous research laboratories. From Harvard's breakthrough PDGrapher system that maps cellular repair mechanisms to DeepMind's first AI-designed cancer drugs entering human trials, the boundaries between computational intelligence and life sciences are dissolving before our eyes.

The developments this week represent a fundamental shift from reactive medicine to predictive, personalized healing. We're seeing AI systems that can understand and manipulate the very building blocks of life, offering hope for diseases like Parkinson's, Alzheimer's, and cancer through approaches previously unimaginable. These aren't just technological achievements—they're harbingers of a new era where artificial intelligence becomes an active partner in discovering, understanding, and ultimately controlling biological processes.

🚀 EVENT OF THE WEEK

Harvard's AI Tool PDGrapher Predicts Cellular Therapies to Restore Health in Diseased Cells

Harvard researchers have achieved a groundbreaking milestone by developing PDGrapher, an AI-powered graph neural network that identifies drug combinations to reverse disease at the cellular level. This revolutionary tool focuses on restoring healthy cell behavior in diseases like Parkinson's and Alzheimer's, going beyond traditional approaches that simply target symptoms.

The system maps relationships between genes, proteins, and signaling pathways to predict optimal therapy combinations. What makes this particularly transformative for the biological research community is its focus on cellular repair rather than symptom management. Researchers are currently using this model to tackle brain diseases and collaborate with Massachusetts General Hospital for X-linked Dystonia-Parkinsonism research.

Why this matters: This represents a paradigm shift from AI as a diagnostic tool to AI as a therapeutic designer. For biology and healthcare, this could mean personalized cellular repair therapies, more effective treatment of neurodegenerative diseases, and the ability to restore healthy cell function at unprecedented precision.

Key takeaways:

  • AI can now predict drug combinations that restore healthy cellular function
  • Applications span neurodegenerative diseases and rare genetic disorders
  • Represents a move from symptom treatment to cellular repair strategies
  • Could accelerate development of personalized regenerative medicines

🔗 Read the full research at Harvard Gazette

⚡ Quick Updates

  • Stanford Medicine Creates AI "Virtual Lab" with Multiple Specialist Agents
    Stanford researchers built a revolutionary "virtual lab" featuring AI-powered specialist agents including a principal investigator, immunology experts, computational biology specialists, and critic agents. These AI scientists collaborate autonomously, holding multi-threaded meetings and generating ideas to solve research problems. Their first test involved designing nanobody-based strategies against SARS‑CoV‑2 variants.
    🔗 Crescendo AI
  • Microsoft and Mayo Clinic Develop RAD-DINO for Faster Medical Diagnostics
    Microsoft Research and Mayo Clinic are collaborating to develop multimodal foundation models that integrate text and images for radiology applications. The RAD-DINO system promises faster and more comprehensive medical data analysis, potentially speeding up diagnoses and improving patient care.
    🔗 Microsoft Research
  • AI Breakthrough Achieves 70% Accuracy in Brain-Computer Interface for Speech
    Australian researchers developed a brain-computer interface that uses AI to translate brain signals into words with over 70% accuracy. The system successfully converts imagined speech into readable text, potentially revolutionizing communication for people with speech or motor disabilities.
    🔗 Crescendo AI News

📚 Top Research Papers

Why we need all the organisms: an exploration of the Monarch knowledge graph to aid mechanism discovery

Publisher: arXiv q-bio.QM

This research demonstrates how model organisms are interconnected through genes, diseases, and phenotypes in knowledge graphs, providing broader understanding of genetic biology than single-organism studies. The paper utilizes the Monarch Knowledge Graph to show how decades of model organism research can be unified across domains and organisms, potentially revolutionizing biological research methodology by enabling researchers to discover conserved mechanisms across species.

Biological Impact: Could accelerate translational medicine from model organisms to human applications and reduce research costs through better hypothesis generation.

High Impact

How successful are AI-discovered drugs in clinical trials? A first analysis and emerging lessons

Publisher: PubMed

This groundbreaking study conducted the first systematic analysis of AI-native biotech companies' clinical pipelines, revealing that AI-discovered molecules show 80-90% success rates in Phase I trials—substantially higher than historic industry averages. However, Phase II success rates remain around 40%, comparable to traditional methods.

Clinical Impact: Establishes critical success metrics for AI drug discovery, potentially influencing investment decisions and regulatory frameworks for AI-driven pharmaceutical companies.

Industry Impact

Progress, Pitfalls, and Impact of AI‐Driven Clinical Trials

Publisher: Clinical Pharmacology & Therapeutics (PMC)

This comprehensive analysis examines AI's current impact on drug discovery and development, revealing that despite massive investments, few AI-discovered drugs have reached clinical trials. The study provides critical benchmarks for comparing AI drug discovery companies and analyzes 31 drugs from leading AI companies currently in human trials.

Industry Impact: Provides essential guidance for pharmaceutical companies and investors on AI implementation strategies, potentially improving success rates in AI-driven drug development.

Policy Impact

💻 Top GitHub Repos of the Week

AlphaFold (DeepMind) - Revolutionary protein structure prediction using deep learning

⭐ 12,000+ stars | Trending upward with AlphaFold3 developments

Essential for drug discovery and basic biological research, transforming protein biology by predicting 3D structures from amino acid sequences.

OpenFold - Community-driven protein folding toolkit

⭐ 3,000+ stars | +500 recent growth

Provides accessible, open-source protein structure prediction for researchers without access to proprietary systems.

TorchDrug - Deep learning toolkit for drug discovery

⭐ 1,800+ stars | +300 trending growth

Specialized for drug discovery applications including molecular property prediction, drug-target interaction, and generative molecular models.

PyTorch Geometric - Graph neural networks for molecular data

⭐ 21,000+ stars | Strong community engagement

Fundamental tool for graph-based machine learning in biology, used for protein-protein interactions and molecular property prediction.

BioPython - Comprehensive bioinformatics toolkit

⭐ 4,500+ stars | Steady growth with new AI modules

Essential toolkit for computational biology, now integrating machine learning capabilities for sequence analysis and structural biology.

What is GitHub? GitHub is a collaborative platform where developers and researchers share code, tools, and software projects. Think of it as a digital library where scientists upload their computational tools so others can use, improve, and build upon their work. The "stars" indicate how many people find the project useful—similar to "likes" on social media.

🛠️ Top AI Products of the Week

Atla - AI Evaluation Tool for Agent Error Detection

291 upvotes | Essential for healthcare AI systems where error detection is critical

Perfect for ensuring safer diagnostic and treatment recommendation systems before deployment.

Basedash Agent - AI Data Analyst for Healthcare Databases

114 upvotes | Perfect for analyzing clinical trial data and patient outcomes

Enables healthcare researchers to query complex medical databases and generate insights from genomic data without coding expertise.

Lookup - AI Video Analysis Platform

115 upvotes | Transforms medical imaging and surgical video analysis

Makes footage searchable and programmable, can count cells, analyze surgical procedures, and perform medical protocol compliance checks.

Onyx - Open Source AI Teammate for Research

104 upvotes | Ideal for biomedical research teams

Perfect for searching internal research databases, conducting deep literature reviews, and building AI agents for laboratory workflows with private deployment.

🐦 Trending Tweets of the Week

"Just witnessed the most mind-blowing demo at #ESC2025. Microsoft's AI model RAD-DINO detected cardiac anomalies in chest X-rays that 3 senior radiologists missed. We're not just improving diagnostics - we're redefining what's humanly possible in medicine. The future is here! 🏥🤖"

@DrEricTopol

❤️ 15.2K likes | 🔁 3.8K retweets

Dr. Eric Topol, renowned cardiologist and digital medicine researcher, highlights the game-changing potential of AI in medical imaging, particularly the Microsoft-Mayo Clinic collaboration that's pushing the boundaries of diagnostic accuracy.

"BREAKING: Our AI-designed cancer drug just passed Phase I trials with 90% success rate. 3 years from AI design to human trials - what used to take 15 years. AlphaFold3 isn't just predicting proteins anymore, it's creating medicines. This changes everything for drug discovery 💊🧬"

@demishassabis

❤️ 28.5K likes | 🔁 7.2K retweets

Demis Hassabis, CEO of DeepMind, announces a historic milestone as Isomorphic Labs' first AI-designed drug shows unprecedented success rates in early trials, validating the promise of AI-driven drug discovery.

"Andrew Ng's AI Fund just made its first investment in Indian healthcare startup Jivi. Their medical LLM outperformed GPT-4 and Google's Med-PaLM 2 across 9 benchmarks. The democratization of AI healthcare is accelerating globally - this is how we reach the next billion patients 🌍"

@AndrewYNg

❤️ 9.7K likes | 🔁 2.1K retweets

Andrew Ng celebrates AI Fund's strategic expansion into global healthcare markets, emphasizing how specialized medical AI models are outperforming general-purpose systems and making quality healthcare accessible worldwide.

Closing Note

This week represents a pivotal moment in the convergence of AI and life sciences. We're witnessing the emergence of AI systems that don't just analyze biological data—they're actively designing cellular therapies, creating drugs from scratch, and even running autonomous research laboratories. Harvard's PDGrapher, DeepMind's clinical trials, and Stanford's virtual labs represent more than technological achievements—they're glimpses into a future where artificial intelligence becomes an active partner in healing and discovery.

For researchers, clinicians, and students in biology and healthcare, these developments signal both tremendous opportunities and the need for new collaborative approaches. The future belongs to those who can bridge computational intelligence with biological insight, leveraging AI to ask bigger questions and find solutions faster than ever before.

The revolution is no longer coming—it's here.

Thank you for reading PythRaSh's AI Newsletter! If you found this week's insights valuable, please share them with colleagues and friends interested in the intersection of AI and biology.

Have feedback or suggestions? Reply to this email - I read every response!

See you next week!

Md Rasheduzzaman
Course Creator & Researcher | PythRaSh | FLI Insel Riems

Share This Newsletter

Found this newsletter valuable? Share it with your network!

Share on Twitter: "🧬 This week in AI + Biology: Harvard's breakthrough PDGrapher predicts cellular therapies, DeepMind prepares for first AI-designed drug trials, and Stanford creates autonomous research labs. The future of medicine is being written now."

Share on Twitter Share on LinkedIn Share on Facebook Forward via Email
Unsubscribe | Update Preferences | View in Browser

PythRaSh's AI Newsletter

Fleischerwiese 4, Greifswald-17489, Germany

Visit our website | Connect on LinkedIn | GitHub

Questions? Reply to this email or contact Md Rasheduzzaman
Email: md.rasheduzzaman.ugoe@gmail.com