5 min read

Agents move into the pathology slide

Agents move into the pathology slide
Nº 01 · The Lede bioRxiv Agents · Infrastructure

Whole-slide pathology goes agent-callable

Whole-slide pathology goes agent-callable
Fig. IbioRxiv, 11 May 2026.

WSInsight wraps single-cell pathology into a cloud-native platform that agents can call as a tool, handling whole-slide image ingest, cell-level segmentation, and downstream queries through one API. The system targets the gap between gigapixel pathology data and the LLM agents (large language model agents — systems that plan multi-step tasks and call external tools) that increasingly drive analysis. By exposing slide-level operations as callable endpoints, it sidesteps the usual bottleneck where pathology pipelines stay walled off from agent workflows. Single-cell resolution at the slide level lands inside the same loop that runs literature search and statistics.

Read the source

ML hunts next-gen IBD targets
Fig. IIbioRxiv, 11 May 2026.
Nº 02 bioRxiv Field report

ML hunts next-gen IBD targets

Harmonized single-cell atlas of inflammatory bowel disease tissue gets paired with an ML target-discovery framework, ranking candidates across patient cohorts rather than within single studies. The harmonization step is the news — cross-study scRNA-seq integration at patient scale has been the rate-limiter on translational target calls, and a working pipeline anchors a new reference workflow for autoimmune target hunts.

Read more
Rare-disease diagnostic agent ships
Fig. IIIarXiv, 11 May 2026.
Nº 03 arXiv Agents · Infrastructure

Rare-disease diagnostic agent ships

A versatile AI agent handles rare-disease diagnosis and risk-gene prioritization in one loop, chaining phenotype matching, variant filtering, and gene ranking. Rare-disease diagnostics has been the canonical case for agentic biomedical AI — narrow, expert-bottlenecked, and data-rich — and a generally-applicable agent here moves the capability frontier past one-disease demos.

Read more
Also Filed · Three Briefs from the queue
Nº 04 arXiv Computational biology

BioTool dataset trains tool-calling LLMs

BioTool releases a tool-calling dataset tuned for biomedical tasks, letting general LLMs learn to invoke domain APIs (BLAST, UniProt, PubMed) without bespoke fine-tunes. Becomes a reference training set for biomedical tool-use — closes one of the gaps between general agents and the lab-instrument integrations real workflows need.

Read
Nº 05 Anthropic Field report

Anthropic turns thoughts into text

Anthropic published research on natural language autoencoders — compressing Claude's internal reasoning into readable text and back. Anchors a new approach to model interpretability; for biology agents whose chain-of-thought drives experimental decisions, auditable reasoning traces become more tractable.

Read
Nº 06 OpenAI Field report

OpenAI lands on AWS

OpenAI's models, Codex, and Managed Agents are now available on AWS, giving enterprises a path to deploy GPT-class agents inside their existing cloud accounts. Removes a procurement and data-residency barrier that had kept many regulated biomedical buyers on the sidelines.

Read

Reply with your discoveries. A human reads them. Forward freely.

Agentic Discovery  ·  Nº Twelve  ·  11 May 2026

Editor's Note

Monday open: cloud-native pathology agents, an IBD target hunt, and a rare-disease diagnostician walk into the lab.

 

Nº 01 · The Lede  —  bioRxiv  —  Agents · Infrastructure

Whole-slide pathology goes agent-callable

Whole-slide pathology goes agent-callable

Fig. I  bioRxiv, 11 May 2026.

WSInsight wraps single-cell pathology into a cloud-native platform that agents can call as a tool, handling whole-slide image ingest, cell-level segmentation, and downstream queries through one API. The system targets the gap between gigapixel pathology data and the LLM agents (large language model agents — systems that plan multi-step tasks and call external tools) that increasingly drive analysis. By exposing slide-level operations as callable endpoints, it sidesteps the usual bottleneck where pathology pipelines stay walled off from agent workflows. Single-cell resolution at the slide level lands inside the same loop that runs literature search and statistics.

Read the source →

Why it matters

Moves digital pathology from a siloed viewer paradigm to a tool-callable surface — pathology joins sequencing and structure prediction as something an agent can drive end-to-end, raising the floor for what a biomedical agent platform is expected to reach.

 

Nº 02  —  bioRxiv  —  Field report

ML hunts next-gen IBD targets

Fig. II  bioRxiv, 11 May 2026.

ML hunts next-gen IBD targets

Harmonized single-cell atlas of inflammatory bowel disease tissue gets paired with an ML target-discovery framework, ranking candidates across patient cohorts rather than within single studies. The harmonization step is the news — cross-study scRNA-seq integration at patient scale has been the rate-limiter on translational target calls, and a working pipeline anchors a new reference workflow for autoimmune target hunts.

Read more →

 

Nº 03  —  arXiv  —  Agents · Infrastructure

Rare-disease diagnostic agent ships

Fig. III  arXiv, 11 May 2026.

Rare-disease diagnostic agent ships

A versatile AI agent handles rare-disease diagnosis and risk-gene prioritization in one loop, chaining phenotype matching, variant filtering, and gene ranking. Rare-disease diagnostics has been the canonical case for agentic biomedical AI — narrow, expert-bottlenecked, and data-rich — and a generally-applicable agent here moves the capability frontier past one-disease demos.

Read more →

 

Also Filed  ·  Three Briefs from the queue

Nº 04  —  arXiv  —  Computational biology

BioTool dataset trains tool-calling LLMs

BioTool releases a tool-calling dataset tuned for biomedical tasks, letting general LLMs learn to invoke domain APIs (BLAST, UniProt, PubMed) without bespoke fine-tunes. Becomes a reference training set for biomedical tool-use — closes one of the gaps between general agents and the lab-instrument integrations real workflows need.

Read →

Nº 05  —  Anthropic  —  Field report

Anthropic turns thoughts into text

Anthropic published research on natural language autoencoders — compressing Claude's internal reasoning into readable text and back. Anchors a new approach to model interpretability; for biology agents whose chain-of-thought drives experimental decisions, auditable reasoning traces become more tractable.

Read →

Nº 06  —  OpenAI  —  Field report

OpenAI lands on AWS

OpenAI's models, Codex, and Managed Agents are now available on AWS, giving enterprises a path to deploy GPT-class agents inside their existing cloud accounts. Removes a procurement and data-residency barrier that had kept many regulated biomedical buyers on the sidelines.

Read →

 

· · ·

Reply with your discoveries. A human reads them. Forward freely.