Industry Guide: Life Sciences

How to Use Claude in Life Sciences: A Guide for Biotech and Research Teams

From the literature to the lab notebook to the filing, with a scientist signing off.

This guide is for biotech scientists, bioinformaticians, clinical operations staff, and regulatory teams who want to put Anthropic's Claude to work on literature review, data analysis, lab documentation, clinical trial protocols, and regulatory writing. It covers the life sciences connectors, skills, and scientific workbench Anthropic has released, and what named companies said in Anthropic's own announcements.

Learning how to use Claude in life sciences means applying it across the R&D chain: searching and synthesizing biomedical literature, writing and debugging bioinformatics pipelines, running quality control on sequencing data, turning lab data into structured records, and drafting clinical trial protocols and regulatory documents from your own evidence. Anthropic has released connectors to tools such as Benchling, PubMed, and ClinicalTrials.gov. Scientists stay responsible for experimental design, data interpretation, and every claim in a filing.

Published

Key Takeaways

  • The strongest life sciences uses are literature synthesis, bioinformatics code, sequencing data QC, lab documentation, protocol drafting, and regulatory writing.
  • Anthropic has released life sciences connectors to Benchling, PubMed, 10x Genomics, Medidata, ClinicalTrials.gov, ChEMBL, and others, plus sample skills for protocol drafting.
  • In June 2026 Anthropic launched a beta scientific workbench with curated skills for genomics, proteomics, structural biology, and cheminformatics.
  • Anthropic states some of its models block professional biology and drug development queries, with a separate verification program for professionals.
  • GxP, data integrity, and regulatory accountability stay with your quality system and your people.

Copy and Paste

Prompts You Can Use Today

Literature synthesis for a target

You are a senior scientist preparing a target assessment. Using the abstracts and papers I have attached, summarize the evidence linking the target to the disease: genetic evidence, expression data, animal models, and any clinical data. For each claim cite the paper by first author and year, rate the strength of evidence as strong, moderate, or weak with a one-line reason, and list contradictory findings separately. Do not cite any paper I did not provide.

Restricting citations to attached papers prevents invented references. Search tools and connectors can gather the papers first.

Single-cell QC script

Write a Python script using scanpy to run quality control on a single-cell RNA-seq dataset in h5ad format. Calculate genes per cell, counts per cell, and mitochondrial fraction; filter using median absolute deviation thresholds that I can set as parameters; save before-and-after violin plots; and write a short report of how many cells were removed at each step. Comment every step and include a small synthetic test.

Set thresholds as parameters so the biologist, not the script, decides what counts as low quality.

Clinical trial protocol section draft

Draft the Study Design and Eligibility Criteria sections of a phase two clinical trial protocol using only the synopsis, investigator's brochure summary, and prior study results pasted below. Follow ICH E6 and ICH M11 structure where applicable, write inclusion and exclusion criteria as testable statements, and mark every design choice that needs medical or statistical sign-off with [REVIEW]. Material: [paste].

The [REVIEW] markers route each judgment call to the right expert.

Regulatory summary from study reports

Using only the nonclinical study report summaries below, draft a nonclinical overview section for an IND submission: pharmacology, pharmacokinetics, and toxicology, with the key findings, doses, and species exactly as reported. Keep a neutral regulatory tone, cross-reference each statement to its study number, and list any inconsistency between reports for the team to resolve. Material: [paste].

Study-number cross-references make the draft checkable by the regulatory writer and QA.

Instrument output to a structured record

Convert the plate reader export below into a structured table with one row per well: plate ID, well, sample ID from the attached plate map, measurement, units, and a flag for wells outside the control range stated in the assay SOP excerpt. Do not change any measured value. Then summarize the run in three sentences for the electronic lab notebook. Export, plate map, and SOP excerpt: [paste].

The instruction not to change measured values protects data integrity.

From Anthropic

What Anthropic offers for Life Sciences

Claude for Life Sciences

Anthropic's life sciences offering, launched in October 2025, with connectors to Benchling, BioRender, PubMed, Wiley's Scholar Gateway, Synapse.org, and 10x Genomics, an Agent Skill for single-cell RNA sequencing quality control, and implementation partners.

· Source

Claude for Life Sciences (January 2026 expansion)

An expansion announced alongside Anthropic's healthcare launch that added connectors to Medidata, ClinicalTrials.gov, ToolUniverse, bioRxiv and medRxiv, Open Targets, ChEMBL, Owkin, and Pathology Explorer, and Agent Skills for scientific problem selection, instrument data conversion to Allotrope, bioinformatics deployment, and sample clinical trial protocol drafts.

· Source

Claude Science

A beta AI workbench for scientists launched in June 2026 for Pro, Max, Team, and Enterprise plans, with more than sixty curated skills and connectors for genomics, single-cell, proteomics, structural biology, and cheminformatics, compute management across local and HPC environments, and a reviewer agent for citation and calculation checks.

· Source

In Practice

How Life Sciences Teams Use Claude

According to Anthropic's published customer story

Schrödinger

In Anthropic's October 2025 life sciences announcement, Schrödinger's CTO and COO said Claude Code turns ideas into working code far faster, in some cases many times faster.

· Read Anthropic's customer story

According to Anthropic's published customer story

10x Genomics

10x Genomics, a connector partner in Anthropic's October 2025 launch, noted that single-cell and spatial analysis traditionally required computational expertise, which the connector aims to make more accessible.

· Read Anthropic's customer story

According to Anthropic's published customer story

Axiom Bio

Axiom Bio's co-founder, quoted in Anthropic's October 2025 announcement, described Claude as invaluable as the company builds AI to predict drug toxicity.

· Read Anthropic's customer story

According to Anthropic's published customer story

Genentech

On Anthropic's life sciences solutions page, Genentech's head of research and early development describes Claude helping reshape day-to-day scientific research and laying groundwork for autonomous labs.

Read Anthropic's customer story

Illustrative walkthrough

From sequencing run to a notebook entry

  1. A bioinformatician asks Claude for a quality control script with thresholds as parameters and a synthetic test.
  2. The script passes the synthetic test, and the bioinformatician runs it on the new dataset.
  3. The biologist reviews the plots and adjusts one threshold that would have removed a rare cell population.
  4. Claude summarizes the run, the thresholds used, and the cells removed for the electronic lab notebook.
  5. The scientist signs the entry and records the final parameters with the data.

Life Sciences and Claude: the Published Numbers

Figures as published by Anthropic, linked to each source.
FigureContextSource
up to 10x faster in some casesSchrödinger on turning ideas into working code with Claude Code, per Anthropic's October 2025 announcementSource
Sonnet 4.5 scores 0.83, against a human baseline of 0.79Protocol QA benchmark reported in Anthropic's October 2025 life sciences announcementSource

Core Workflows

Claude Life Sciences Use Cases Across R&D

These are the places where AI for life sciences and AI for biotech teams saves the most scientist time.

Literature review

Biomedical literature review and target assessments built from papers you supply or retrieve through connectors such as PubMed and bioRxiv, with every claim cited.

Bioinformatics

AI for bioinformatics covers writing and debugging Python and R, sequencing QC, and pipeline work in tools such as Nextflow and scVI-tools.

Lab data and documentation

Instrument outputs turned into structured records, experiment summaries for the lab notebook, and SOP drafts, with measured values left untouched.

Clinical trial protocols

Clinical trial protocol drafting from synopses and prior data, with eligibility criteria written as testable statements and judgment calls marked for review.

Regulatory affairs

AI for regulatory affairs in biotech means drafting IND sections, responses to agency questions, and summaries from study reports, cross-referenced for QA.

Discovery support

AI for drug discovery work here means organizing target evidence, compound data from sources such as ChEMBL, and hypotheses for scientists to test.

What Anthropic Released

What Lab Tools Does Claude Connect To?

Anthropic's October 2025 life sciences launch added connectors to Benchling, BioRender, PubMed, Wiley's Scholar Gateway, Synapse.org, and 10x Genomics, alongside general tools such as Google Workspace, Microsoft 365, Databricks, and Snowflake. It also published an Agent Skill for single-cell RNA sequencing quality control.

In January 2026 Anthropic expanded that set with connectors to Medidata, ClinicalTrials.gov, ToolUniverse, bioRxiv and medRxiv, Open Targets, ChEMBL, Owkin, and Pathology Explorer. It added Agent Skills for scientific problem selection, converting instrument data to the Allotrope format, bioinformatics deployment with scVI-tools and Nextflow, and a sample skill for clinical trial protocol draft generation.

In June 2026 Anthropic launched a scientific workbench in beta for Pro, Max, Team, and Enterprise users, with more than sixty curated skills and connectors preconfigured for genomics, single-cell work, proteomics, structural biology, and cheminformatics. According to Anthropic it renders protein structures and genome browser tracks, manages compute on laptops, clusters, or GPUs, and includes a reviewer agent that checks citations and calculations.

Clinical and Regulatory

How Is AI Used in Clinical Trials and Regulatory Work?

AI for clinical trials is mostly about documents and data flow. Protocols, amendments, informed consent forms, site communications, and study reports are long, structured, and heavily cross-referenced. Claude drafts sections from your synopsis and prior data, checks internal consistency, and turns eligibility criteria into testable statements. With a connector such as ClinicalTrials.gov, it can also help compare your design with registered studies.

Can AI draft an IND filing? It can draft sections from your study reports, and Anthropic has published a sample skill for clinical trial protocol drafts. It cannot decide what goes into a submission, and it does not replace your regulatory strategy or quality review. Every statement must trace to a controlled source document, and every draft moves through your document control process like any other.

Regulatory writing benefits from narrow prompts: one section, one set of source documents, strict cross-references, and markers for anything needing expert sign-off.

Choosing Tools

Claude vs ChatGPT for Life Sciences

The Claude vs ChatGPT for life sciences comparison is best run on your own work: a literature question you already answered, a script you already wrote, and a protocol section already approved. Compare accuracy, citation behavior, code quality, and how each handles uncertainty.

For most organizations, the deciding factors are the connectors to your scientific systems, data terms for unpublished research, validation requirements under GxP where applicable, and what your IT and quality teams approve.

One practical note: Anthropic has said its Fable models will continue to block professional biology and drug development queries, and it has announced a verification program with the U.S. government for life sciences professionals to access more capable models. Check which models your plan provides and what applies to your work.

Getting Started

A First Month for a Biotech Team

Start with literature

Use a question your team already answered. Check every citation against the paper before trusting the workflow.

Add analysis code

Have Claude write scripts with tests on synthetic data, then run them on real data only after review.

Connect your systems

If your organization has enabled connectors such as Benchling or PubMed, use them so work stays grounded in your records.

Agree the GxP boundary

Decide with quality which work is exploratory and which is regulated, and document how AI-assisted regulated work is reviewed.

Limits

Scientific Integrity and Where Scientists Decide

Claude can produce a plausible citation that does not exist, misstate a dose, or write code that runs and still computes the wrong thing. In science that is not a small error. Restrict citations to papers you provide or retrieve, check every number against the source, and test code on data where you know the answer.

Experimental design, interpretation of results, and the decision to advance a program belong to scientists. Claude can suggest hypotheses and point out gaps, but the evidence standard is set by your team and, for clinical and regulatory work, by regulators.

Unpublished data, patient-level clinical data, and proprietary compounds are highly sensitive. Use only approved deployments, follow your data governance and patient privacy rules, and keep regulated records inside validated systems.

Bring Rigor to Your AI Workflows

Scientists who get reliable results from AI have practiced grounding it in sources and checking its output. The free Nightschool AI curriculum is hands-on from the first lesson.

Frequently Asked Questions

How is AI used in clinical trials?

Mostly for documents and data flow: drafting protocol sections and amendments, writing eligibility criteria as testable statements, checking consistency across documents, and summarizing study data. Anthropic has released connectors to Medidata and ClinicalTrials.gov and a sample skill for protocol drafts. Medical, statistical, and regulatory decisions stay with the study team.

Can AI draft an IND filing?

AI can draft sections of an IND from your study reports, such as nonclinical overviews, with cross-references to source documents. It cannot decide the regulatory strategy or replace quality review. Treat every AI-drafted section as a controlled draft that moves through your normal review and approval process.

Can AI draft a clinical trial protocol?

Yes, as a first draft from your synopsis and prior data. Anthropic has published a sample Agent Skill for clinical trial protocol draft generation. Every design choice, from endpoints to eligibility criteria, needs medical and statistical sign-off before the protocol is final.

What lab tools does Claude connect to?

Anthropic has released connectors to Benchling, BioRender, PubMed, Wiley's Scholar Gateway, Synapse.org, 10x Genomics, Medidata, ClinicalTrials.gov, ToolUniverse, bioRxiv and medRxiv, Open Targets, ChEMBL, Owkin, and Pathology Explorer, among others. Availability depends on your plan and your organization's setup.

What tools has Anthropic released for life sciences?

Anthropic launched a life sciences offering in October 2025 with connectors and a single-cell QC skill, expanded it in January 2026 with clinical and discovery connectors and more skills, and launched a beta scientific workbench in June 2026. Details and dates are in the offerings section of this page, linked to Anthropic's announcements.

What is Anthropic's scientific workbench?

Launched in beta in June 2026, it is an environment for scientists with more than sixty curated skills and connectors for areas such as genomics, proteomics, structural biology, and cheminformatics. Anthropic says it renders protein structures and genome tracks, manages compute across laptops, clusters, and GPUs, and uses a reviewer agent to check citations and calculations.

How does AI speed up drug discovery?

In practice, mainly by reducing the time scientists spend searching literature, writing analysis code, and organizing data from sources such as ChEMBL and Open Targets. Anthropic's announcement quotes Schrödinger saying Claude Code turns ideas into working code much faster. Whether that shortens a program depends on the science.

Is it safe to use AI with unpublished research data?

Only through deployments your organization has approved, with data terms your IT, legal, and quality teams have reviewed. Patient-level clinical data carries additional privacy obligations. Regulated records should stay in validated systems.

Spend More Time on the Science

Hands-on lessons that respect how scientists work.

Nightschool AI is an independent learning platform and is not affiliated with, endorsed by, or sponsored by Anthropic. Claude is a trademark of Anthropic, PBC. Looking for Anthropic's official Claude Academy? It's at academy.claude.com.