BenchSci

BenchSci is a biotechnology artificial intelligence company that develops research technology to help pharmaceutical and biotechnology organisations understand disease biology and improve the drug discovery process. Founded in 2015 and headquartered in Toronto, Canada, the company focuses on reducing the time and uncertainty involved in preclinical research. Its technology brings together scientific publications, experimental evidence, biological data, computational models, and AI-powered research workflows to help scientists make better-informed decisions during early-stage drug development.

The company was established after its founder, Tom Leung, experienced a problem with antibody selection during his doctoral research. A poorly selected antibody contributed to the loss of valuable patient samples, even though relevant experimental evidence existed somewhere in scientific literature. This experience highlighted how difficult it could be for researchers to find and evaluate the information needed to plan reliable experiments. Tom Leung subsequently worked with machine learning researchers David Chen and Elvis Wianda to explore whether AI could read scientific publications and extract the experimental information researchers needed. Liran Belenzon later joined the founding team as CEO.

BenchSci initially concentrated on AI-assisted antibody and reagent selection. Its first product helped researchers identify suitable laboratory reagents by analysing experimental evidence in published scientific papers. This gave scientists a way to evaluate antibodies and other research materials using evidence from previous experiments rather than relying exclusively on product descriptions or manual literature searches. The original product helped establish the company’s focus on applying machine learning to scientific information and improving the reliability of laboratory research.

Over time, BenchSci expanded beyond reagent selection into a broader research intelligence platform. Its technology developed to support the investigation of disease biology, target identification, experimental planning, and the evaluation of biological relationships. The company introduced ASCEND in January 2023, an AI-powered platform designed to help preclinical researchers explore disease biology and connect evidence from scientific literature with experimental data. ASCEND represented an important development because it moved BenchSci from helping researchers select individual laboratory materials towards supporting wider scientific decision-making.

BenchSci’s current platform is EMET, an agentic research environment developed for pharmaceutical and biotechnology research and development. EMET brings together scientific data, AI models, specialised research skills, software tools, and workflows in a unified environment. It is designed to help scientists investigate complex research questions, generate hypotheses, identify biological targets, design experiments, evaluate potential risks, and improve lead optimisation. Rather than functioning only as a conventional literature search engine, EMET uses AI orchestration and scientific reasoning to organise multi-step research tasks.

One of EMET’s central capabilities is its ability to interpret complex scientific questions written in natural language. Researchers can ask questions about a disease, biological pathway, molecular target, experimental method, or potential therapeutic opportunity. The platform can break a broad question into smaller research tasks, identify suitable scientific resources, and coordinate specialised AI skills to investigate different parts of the problem. This allows researchers to explore relationships across several sources of evidence without manually conducting every search and comparison themselves.

The platform draws on a large collection of scientific information. BenchSci describes access to more than 38 million scientific publications, including 16 million closed-access papers available through publisher partnerships and licensing arrangements. Its information sources also include patents, preprints, omics information, clinical trial records, and more than 1,000 additional databases. Bringing these resources together helps researchers investigate biological questions using evidence from different areas of scientific literature and experimental research.

A major part of BenchSci’s technology is its proprietary scientific knowledge graph. This graph connects information about diseases, biological targets, pathways, experimental methods, reagents, and research models. It is designed to make relationships between scientific concepts easier to explore. Instead of treating every publication as an isolated document, the system organises relevant information into connected relationships that researchers can investigate in the context of a particular disease or research programme.

BenchSci also maintains curated biological and experimental datasets. Its website describes collections covering antibodies, RNA interference records, CRISPR information, cell lines, and animal models. These resources help researchers examine experimental evidence and identify relevant biological materials for their work. The company says its data is curated by a team of PhD scientists, adding scientific expertise to the process of organising information from different sources.

EMET includes more than 100 specialised scientific skills designed to support different stages of preclinical research. These skills cover activities such as target identification, target validation, omics analysis, biomarker development, and experimental design. The platform can coordinate these capabilities to investigate complex questions and produce research outputs supported by traceable evidence. Its approach is intended to help scientists spend less time connecting disconnected databases and more time evaluating scientific findings.

Another important capability is hypothesis generation. Drug discovery often begins with questions about the biological mechanisms responsible for a disease and whether a particular target could be useful for developing a treatment. Researchers need to understand existing evidence, identify gaps in scientific knowledge, and evaluate whether a proposed relationship is sufficiently supported. EMET helps organise this information so that scientists can investigate potential targets, examine biological connections, and develop hypotheses for further experimental validation.

Target identification and validation are also important parts of the platform. Researchers can investigate how a biological target relates to a disease, which pathways it influences, what experimental evidence supports the relationship, and what uncertainties remain. These activities can help research teams compare potential targets and identify areas that deserve additional investigation. The platform is intended to support scientific evaluation rather than replace the experimental work required to establish whether a target is suitable for drug development.

EMET also supports experimental design. Once researchers have developed a hypothesis, they need to determine how best to test it. This may involve selecting suitable cell lines, animal models, antibodies, genetic perturbations, or experimental methods. BenchSci’s research environment helps scientists investigate relevant evidence and identify options for planning experiments. By connecting scientific publications with curated experimental information, it can help teams make more informed choices before committing laboratory resources.

Risk assessment is another part of the research workflow. Drug discovery programmes can encounter problems when researchers overlook important biological relationships, safety concerns, or limitations in the available evidence. EMET is designed to help identify potential risks and examine the scientific information associated with a proposed research direction. This can support earlier investigation of questions relating to biological feasibility, experimental reproducibility, and possible safety or efficacy concerns.

The platform also supports lead optimisation, a stage of drug discovery in which researchers investigate and improve promising candidate molecules. This work requires an understanding of the relationship between molecular properties, biological activity, and experimental results. EMET brings scientific information and computational capabilities together to support research teams as they evaluate potential leads and consider the next steps in development.

BenchSci’s agentic architecture allows different AI capabilities to work together on a research task. The system can coordinate frontier AI models with specialised scientific tools and computational resources. Its technology stack includes models and tools such as ESM-2, AbLang2, and RDKit, which support different aspects of biological sequence analysis, antibody-related research, and chemical informatics. The model-agnostic approach is intended to let BenchSci incorporate suitable new models as they become available rather than relying permanently on one foundation model.

Scientific traceability is an important part of the platform’s design. Research decisions need to be supported by evidence that scientists can inspect and evaluate. EMET aims to provide explainable outputs with connections to the underlying publications, datasets, and scientific relationships used in its reasoning process. This helps researchers assess where an insight came from and decide whether it is relevant to their particular research question. The platform’s evaluation approach also includes scientific testing and validation across multiple benchmarks.

BenchSci’s enterprise offering is designed for pharmaceutical and biotechnology organisations that need research technology connected to their internal information and established workflows. Companies can incorporate proprietary data, internal research records, and other information that may not be available in public scientific databases. EMET is also designed to support custom integrations and connections to laboratory automation environments, including lab-in-the-loop and self-driving laboratory workflows. These capabilities can help organisations connect computational research with experimental activities.

A dedicated scientific team is another part of BenchSci’s enterprise approach. The company describes deployments that include PhD-level scientific support, training, change management, and custom connector development. This is intended to help organisations adapt the platform to their research programmes, therapeutic areas, and internal processes. Such support is relevant to larger research teams that need more than access to a general-purpose software interface.

BenchSci has also developed ASCEND, which remains an important part of its product history and scientific research ecosystem. ASCEND was designed to help scientists investigate disease biology through specialised research applications. Its tools included Selector, Navigator, Architect, Defender, Conversational AI, and project summarisation capabilities. These functions supported activities such as reagent selection, biological target investigation, experiment planning, risk evaluation, and the organisation of research findings. The company’s newer EMET environment brings its scientific data and AI capabilities into a broader agentic research workflow.

The original Selector application is particularly relevant to antibody and reagent research. It helps scientists evaluate commercially available research materials by examining published experimental evidence. This can support decisions about which antibodies, cell lines, and other biological models may be appropriate for a particular experiment. The approach is valuable because reagent selection can affect experimental reliability, reproducibility, and the interpretation of research results.

BenchSci’s customers and research partners include major pharmaceutical and biotechnology organisations. The company has described adoption across leading pharmaceutical companies and research institutions, including collaborations involving organisations such as Amgen. These relationships have helped shape its enterprise research technology and demonstrate its focus on the needs of large-scale preclinical research teams.

The company has also attracted investment from venture capital organisations, including Generation Investment Management, Google Gradient Ventures, F-Prime Capital, TCV, and Inovia Capital. Its funding has supported the development of its scientific data infrastructure, AI models, research platform, and enterprise capabilities.

BenchSci is primarily intended for scientists and research organisations working in pharmaceutical and biotechnology research. Its platform is particularly relevant to preclinical researchers, computational biologists, molecular biologists, translational researchers, and drug discovery teams. Pharmaceutical companies can use it to investigate biological targets and evaluate research programmes, while biotechnology companies can use its scientific intelligence capabilities to support early-stage discovery. Academic and nonprofit researchers can also access selected parts of the company’s research technology through its academic offering.

The organisation’s work is focused on improving the infrastructure behind scientific discovery. Research teams often need to combine information from scientific publications, patents, internal datasets, experimental records, and specialist software. These resources may use different formats and terminology, making it difficult to develop a complete understanding of a biological question. BenchSci aims to reduce this fragmentation by connecting scientific evidence, curated data, AI models, and research workflows in one environment.

BenchSci is not a general-purpose AI chatbot or a laboratory equipment manufacturer. It is a specialised research technology company that provides AI-powered scientific intelligence and research infrastructure. Its platform supports scientists in investigating biological questions and planning research, while experimental validation and professional scientific judgment remain essential. The value of its technology lies in helping research teams access, connect, and evaluate complex scientific information more efficiently.

Pricing

BenchSci offers different access arrangements for academic and industry users.

• Academic plan – Free: Limited access for eligible academic, nonprofit, and government researchers. Access includes selected research capabilities and is subject to academic-use terms.
• Enterprise plan – Custom pricing: Designed for pharmaceutical and biotechnology organisations. Includes expanded data access, closed-access publications, knowledge graph capabilities, custom integrations, enhanced research workflows, security features, and dedicated support.
• Free trial / pilot: Prospective enterprise customers can begin with a free pilot or trial arrangement, subject to the company’s access process.

Enterprise costs depend on the organisation’s research requirements, selected capabilities, data needs, integrations, and deployment arrangements.

Review

BenchSci is a specialised AI research platform with a strong focus on improving the quality and efficiency of pharmaceutical and biotechnology research. Its main strength is the way it connects scientific publications, biological datasets, experimental evidence, and AI-powered workflows. This gives researchers a more organised environment for investigating complex biological questions and planning preclinical experiments.

EMET is a significant development because it moves beyond traditional scientific search. Its agentic approach can coordinate different research skills, models, and information sources to investigate multi-step questions. This is useful for scientists who need to examine biological targets, understand disease mechanisms, generate hypotheses, and evaluate experimental options. The ability to work with natural-language research questions also makes the platform more accessible than workflows that depend entirely on manually searching multiple databases.

The scientific knowledge graph and curated datasets are other major strengths. Connecting information about diseases, pathways, targets, antibodies, cell lines, and experimental models can help researchers discover relationships that may be difficult to identify through isolated searches. The company’s publisher partnerships also provide access to licensed scientific publications, which can expand the evidence available for research investigations.

BenchSci’s scientific focus is another advantage. Its platform is designed around the needs of preclinical research rather than adapting a general-purpose chatbot for biomedical work. Specialised scientific skills, curated information, and connections to tools such as ESM-2, AbLang2, and RDKit make it relevant to research teams working with complex biological and chemical data.

The enterprise offering is particularly useful for large pharmaceutical and biotechnology organisations. Custom integrations, internal data support, dedicated scientific teams, and laboratory workflow connections can help companies adapt the platform to their research environments. The free academic offering also provides selected access for eligible academic, nonprofit, and government researchers.

However, BenchSci is not designed for every type of user. Its main focus is biopharma research and preclinical drug discovery, so people looking for general academic writing, everyday productivity, or broad educational AI tools may find it too specialised. Its more advanced enterprise capabilities are also intended for organisations with established scientific research programmes.

Another consideration is that enterprise access involves a commercial relationship and may require implementation planning. Organisations may need to evaluate data requirements, security arrangements, internal integrations, and staff training before deploying the platform across research teams. The custom nature of these deployments can make adoption more involved than signing up for an ordinary self-service application.

AI-generated research insights also require scientific evaluation. Although EMET is designed to provide traceable and evidence-supported outputs, researchers must still examine the underlying studies, assess the quality of experimental evidence, and validate important hypotheses through appropriate scientific methods. The platform should support scientific decision-making rather than replace experimental validation or expert judgment.

BenchSci is a strong option for pharmaceutical and biotechnology research organisations that need advanced scientific intelligence and AI-supported preclinical workflows. Its combination of EMET, scientific data infrastructure, curated biological information, specialised research skills, and enterprise support makes it particularly relevant to teams investigating disease biology and developing potential medicines.