Bohr Science Space Automates Research ‘Heavy Lifting’ to Free Scientists’ Time

1. From Scientific Question to Full‑Process Automation

Scientists often waste valuable time on searching, organizing, validating, reproducing, and communicating results. Bohr Science Space (public beta) offers a unified desktop environment where researchers pose questions and AI scientists (SciMaster) handle the “heavy lifting.” The system automatically retrieves papers and patents, parses text, figures, molecules, and formulas, and produces structured research reports that highlight consensus, controversies, and knowledge gaps.

When a hypothesis shows promise, the AI composes executable experiment designs, specifying variables, technical routes, key steps, and result‑evaluation criteria. Researchers can iteratively request next steps without switching tools.

Domain‑Specific AI Scientists

BioMaster : Executes reproducible bio‑informatics pipelines—linking literature, GEO data, and code to reproduce analyses, handling data acquisition, environment setup, debugging, and result verification for tasks such as paper‑reproduction, single‑cell analysis, and gene‑variant interpretation.

PharmMaster : Drives drug‑discovery workflows from target research, patent landscape, SAR analysis, molecular design, to ADMET evaluation, exemplified by an ITK inhibitor project that automates data collection, patent analysis, SAR interpretation, molecular screening, co‑crystal analysis, retrosynthetic planning, and activity prediction.

MatMaster : Supports materials‑science pipelines—organizing literature, designing candidate systems, running DFT or molecular dynamics simulations, and recommending experimentally verifiable material directions.

2. Why AI Can Execute Real Scientific Tasks

The platform connects over 200 million papers and patents (Science Navigator), 50 000 scientific computing tools (DeployMaster) with large models such as DeepScience‑YuZhi, Uni‑Mol, Uni‑AIMS, and computational tools like Uni‑FEP and Uni‑Dock. In the wet‑lab stage, Uni‑Lab‑OS links more than 150 instrument categories (≈1800 devices), forming a closed‑loop research chain.

Uni‑Parser extracts multimodal scientific information (text, formulas, tables) with >98% accuracy and identifies visual elements (charts, molecular structures) for machine‑readable conversion. The Knowledge‑Graph‑based Large Knowledge Model (LKM) organizes entities, evidence sources, and research context, enabling AI to understand consensus, disputes, and open questions.

The SciX framework decomposes research goals into executable steps, orchestrates knowledge, models, software, compute, and lab resources, and performs checkpoint verification. Its Sandbox provides isolated, recoverable environments across heterogeneous compute resources, achieving a 98.8% success rate over ~1.85 million task launches.

3. Benefits for Individuals, Teams, and Organizations

Researchers save time on database queries, environment configuration, tool switching, and result collation, allowing a problem‑driven workflow.

Project groups obtain traceable records of methods, parameters, and outcomes, facilitating review, reuse, and iteration.

Institutions and enterprises gain a unified platform to connect internal data, compute, specialized tools, and domain expertise, supporting human‑AI collaboration on real tasks.

4. Bohr 107 Initiative

To prioritize high‑impact scientific problems, DeepScience launches the “Bohr 107” program, inviting global researchers to submit valuable questions. Selected problems receive end‑to‑end support from Bohr Science Space, from literature mining to hypothesis testing and validation.

5. Returning Time to Scientific Creation

AI scientists excel at repetitive, indispensable tasks, freeing human scientists to pose deeper questions, interpret mechanisms, and explore uncharted phenomena. The platform’s self‑learning agents continuously evolve from executed workflows, improving future assistance.