Search

Agentic AI for Immune-Related Adverse Event Detection in Cancer Immunotherapy - OncoDaily

An agentic AI system using specialized pipelines and an LLM assesses immune-related adverse events from cancer immunotherapy, extracting presence, status, severity, attribution to therapy, and clinician certainty, with evidence links for auditability. Evaluated across retrospective, prospective, and randomized studies, it speeds up and standardizes irAE curation but struggles with precise severity assessment.

Prof. Dr. Günal Explained the Vital Differences Between Cold and Coronavirus (COVID-19)

Translated title. Original: Prof. Dr. Günal Grip ve Soğuk Algınlığı Arasındaki Hayati Farkları Açıkladı

An infectious-disease expert in Samsun distinguishes flu from cold, warning about higher risks for vulnerable groups, noting a recent rise in flu-like cases with seasonal changes and emphasizing vaccination and hygiene as key protections.

GPT-5 Unravels the Mystery of T Cells | The Latest Case of AI × Immunology | Become AI-Savvy

Translated title. Original: GPT-5がT細胞の謎を解明｜AI×免疫学の最新事例 | AIに強い人になろう

The article reports that GPT-5 Pro helped immunologist Derya Unutmaz solve a three-year research problem about T-cell behavior, with possible relevance to cancer and autoimmune disease research. It stresses that experiments and human review are needed before AI findings can be trusted or applied in medicine.

AI Digital Twins Identify Drugs to Prevent Early Labour

Stanford Medicine researchers developed SIMCO, an AI system that creates digital twins of pregnant women's immune systems to test drug effects virtually. The study, published in Science Advances, suggests aspirin combined with lansoprazole could significantly delay labour, with folic acid also potentially slowing immune changes related to labour; these findings aim to reduce preterm birth risk through AI-assisted drug repurposing.

AI Designs 35,000 Antibodies in Weeks, Opening New Possibilities for Food Allergy Treatment - SnackSafely.com

Anthropic, via Claude, designed ~35,000 antibody candidates in a month with Divergence Labs to target IgE for food allergies, showing several designs that bind or detach IgE in lab tests; no human trials yet, but the approach promises faster discovery than traditional methods.

AI Designs 35,000 Antibodies in Weeks, Opening New Possibilities for Food Allergy Treatment - SnackSafely.com

Anthropic, via its Claude AI system, designs thousands of antibody candidates to disrupt IgE-mediated food allergies, with some candidates showing promising lab activity compared to Xolair. While results are fast and suggest potential, no human trials exist yet, and safety and efficacy remain unproven; the effort spans manufacturing and regulatory applications with multiple industry partners.

How AI Regulation Works: Rules, Risk Tiers and Oversight

Regulation of AI combines laws, guidance, oversight, and international cooperation, tailored to the system, actors, harms, and jurisdiction. Different regions (EU, US, UK, standards like NIST) use varying power and timing; regulation targets specific uses and actors rather than imposing a universal license.

ASCII.jp: Miyu Moriyama: Researcher unraveling the mystery of mucosal immunity through intranasal vaccines

Translated title. Original: ASCII.jp：森山美優：経鼻ワクチンを手がかりに粘膜免疫の謎を解く研究者

Miyu Moriyama, a Yale researcher, investigates intranasal vaccines to prevent infection, using animal studies to test efficacy and studying mucosal immunity to explain their function.

Can AI Be Regulated Effectively? EU and U.S. Rules Explained

Regulation can reduce AI harm if rules cover applicable systems, require evidence of risk management, empower regulators, and adapt over time; current EU and US approaches have not yet shown proven real-world harm reduction, and success requires translating broad goals into enforceable duties.

[2610.33195] Agent-guided epitope excision enables economical de novo binder design

The study reports that trimming targets to epitope windows can reduce protein-design compute while preserving binding interfaces. Agent-guided workflows improved computational hit rates, and yeast-display tests identified nanomolar binders.