BenchSci: Reimagining Pre-Clinical Drug Discovery

Welcome to the 6th Network Effects Newsletter.

If you're new here, this newsletter is all about unpacking the vision, strategy, and execution behind the world’s leading tech companies.

Today, we’re exploring BenchSci, which was founded in 2015 by Tom Leung (CSO), David Chen (CTO), Elvis Wianda (CDO), with Liran Belenzon (CEO) joining in 2016

Let’s dive in.

📝 Overview 

BenchSci is an AI-powered reagent intelligence platform that transforms biomedical literature into experiment-specific insights. BenchSci focuses on antibodies, representing 88% of all research reagents used in preclinical drug development. Given that 30-40% of drug discovery costs are concentrated in preclinical R&D, BenchSci addresses a critical pain point by enabling more efficient, evidence-based experimentation.

The company’s flagship product, ASCEND, is a machine learning platform that decodes complex scientific text and figures to surface actionable insights across millions of published experiments. ASCEND is currently used by over 49,000 scientists worldwide, supporting key functions such as target identification, biomarker discovery, and hypothesis generation.

Within BenchSci’s top 10 enterprise accounts, ASCEND has driven measurable R&D efficiency gains: users uncovered novel targets or disease indications in 22% of key projects, while reducing unnecessary experimentation by 40%, materially accelerating time-to-insight in the preclinical research cycle.

📌 Thesis 1 - Markets Ripe for Innovative Disruptors 

Biopharma’s innovation engine is fundamentally broken. Over the past three decades, R&D costs have increased more than 6x while new drug approvals have remained flat, a paradox often referred to as Eroom’s Law (Moore’s Law in reverse). 

Until recently, clinical operations (e.g., trial design, patient recruitment) have been the focal point for digital transformation, platforms like Medidata or Veeva dominate that segment. In contrast, preclinical research has remained largely manual and fragmented: scientists sift through thousands of papers, manually compare reagent options, and design experiments in isolation. ASCEND is among the first true horizontal AI infrastructures purpose-built for preclinical workflows, encompassing target ID, reagent selection, experimental planning, and risk assessment. By automating literature mining, extracting structured insights from complex figures, and surfacing context-specific reagent recommendations, BenchSci demonstrably cuts experimental planning time by 30–50% and reduces failed experiments by up to 40%.

Furthermore, there is a growing competitive pressure from emerging BioPharma (EBP) to compete with incumbents. It creates an appetite for BioPharmas to invest in alternative platform solutions to obtain an edge over their competitors. 

Adopters of the ASCEND platform have shown real value through reduced unnecessary experimentation and an increase in productivity. For example, Novartis has saved over $14M by deploying BenchSci by avoiding unproductive experimental dead ends.

📌 Thesis 2 – Data Flywheel Translate to Lower Error Rates

Over the past decade, BenchSci has formed a running virtuous cycle of data partnership with 200+ vendor catalogues and private publisher content (e.g., Springer, Wiley, Oxford), covering over 10M+ scientific publications, 47,000+ validated independent datasets and 6.7M+ reagents.

This rich, proprietary data set is not easily replicable and forms the core of BenchSci's competitive edge. Beyond that, BenchSci creates a dynamic flywheel, where this proprietary data fuels BenchSci's ability to deliver superior insights, driving increased platform usage and stickiness among researchers. This influence further encourages these publications to partner with BenchSci, enriching the platform with valuable content, attracting more users, and enhancing its overall value in a self-reinforcing loop.

Layering on customers’ proprietary internal data (e.g., Electronic Lab Notebooks, reagent purchasing histories) further personalizes and enriches the platform, increasing switching costs and embedding ASCEND deeper into R&D workflows.

While broader AI drug-discovery platforms (e.g., Recursion, Insitro) are moving upstream, their primary focus remains on molecule generation and late-stage optimization. BenchSci’s specialization in experimental design and reagent intelligence has allowed it to build deep expertise in preclinical R&D, with deep network effects from the usage of proprietary data and the performance of its intelligence platform. BenchSci can maintain its leadership position, even as horizontal players expand into the preclinical space.

🌱 Genesis Story

In 2008, the U.S. government mandated the NIH Public Access Policy, requiring all NIH-funded research to be freely available through PubMed Central. By 2015, this repository had amassed over 1.2 million peer-reviewed studies, however, the vast majority of research findings remained buried, inaccessible to traditional methods of search and interpretation due to the sheer volume and the complexity of scientific figures and languages

Recognizing this gap, Tom Leung, then a PhD candidate at the University of Toronto, had an idea to leverage AI to learn from scientific papers and help build, test and validate solutions to the antibody reproducibility crisis. To build this platform, Leung recruited David Chen and Elvis Wianda, fellow researchers with deep expertise in data science and AI. 

In 2016, Liran Belenzon, who was working for the Creative Destruction Lab (CDL), a startup accelerator focused on Machine Learning, reached out to recruit to BenchSci team to apply for the accelerator program. The team was successful in joining CDL, and Liran started working with them to support their commercialization, then joined as co-founder and CEO. 

In 2017, BenchSci launched its first application, AI-Assisted Antibody Selection, to help scientists reduce experimental failure. The solution used both experiment-specific text ML and proprietary vision ML models that could understand the type of experiment being conducted. Next, our team expanded the technology to guide the selection process of many reagents.

🖥️ Products & Services 

BenchSci’s ASCEND platform offers a specialized suite of AI-powered solutions designed to dramatically improve preclinical R&D workflows for global pharmaceutical companies. The platform can be broken down into three core components: Application, Technology and Data

Applications

The ASCEND platform consists of 4 applications, including Selector, Navigator, Architect, and Defender. This suite of applications empowers users to discover biological connections, surface contextual experimental evidence, and uncover safety and efficacy risks in experiments early to move the most promising projects forward faster. 

• Selector: Selector helps users identify the right reagents and model systems for experiments by leveraging insights from published literature. This application is the starting point in the research process, providing scientists with relevant and high-quality data to make informed decisions on reagents and model systems. By searching across commercial vendors and accessing literature-based insights, scientists can quickly narrow down their options and choose the most supported products.

• Navigator: Navigator is used for exploring published data to uncover biological connections, such as relationships between proteins, pathways, or diseases. It enables users to surface relevant experimental evidence and prioritize research targets based on what is already known in the scientific field.

• Architect: Architect is a tool for designing experimental workflows. It consolidates the reagents and model systems chosen in Selector, integrates research objectives, and allows for planning experiments based on previously generated data from Navigator. It also ensures that the experiments are designed for reproducibility, speed, and regulatory compliance.

• Defender: Defender operates as a safety net, identifying potential risks in preclinical experiments and flagging concerns about downstream clinical relevance. It helps assess whether the experiment might face issues related to efficacy or safety before proceeding, making it a crucial tool in ensuring that research is heading in the right direction.

Technology 

The underlying technology that BenchSci leverages to provide insights to researchers is through state-of-the-art machine learning algorithms with an extremely high level of precision. From ingestion to interpretations to indexing complex scientific literatures with experimental figures into a searchable knowledge base for application uses. 

In addition, BenchSci develops specialized AI assistants designed to deeply understand pharmaceutical workflows and deliver scalable, explainable value to biopharmaceutical R&D. These AI assistants are built to integrate seamlessly into complex preclinical and clinical research workflows, across built five key dimensions that each AI assistant targets:

1. Scientist Persona: The AI assistants are tailored to the specific roles and expertise of scientists within an organization, from scientists to project leaders.

2. Task Type: The platform is designed to support research activities such as literature review, hypothesis generation, experimental design, or risk assessment.

3. Therapeutic Area (TA): BenchSci’s assistants are equipped with domain-specific knowledge across diverse therapeutic areas

4. Development Stage: The AI platform is designed to adapt to the different stages of the drug development pipeline. Each AI assistant offers insights that are contextually relevant to the stage of development

5. Return on Investment (ROI): A unique feature of the platform is its focus on ROI. The AI tools are designed to provide users with insights that not only help move research forward but also optimize resource allocation. 

Data Sources 

The ASCEND platform is built on a robust knowledge graph and ontological knowledge base, encompassing over 10 million scientific publications and more than 6.7 million antibodies. These data points form a dynamic map of disease biology, integrating both internal and external sources to offer a comprehensive view of scientific references.

BenchSci’s Database 

BenchSci’s database contains over 27 million peer-reviewed scientific publications, abstracts, patents, and internal client data, enriched with third-party databases for a thorough understanding of disease biology, scientific experiments, and preclinical research

It includes both publicly available scientific literature published on PubMed Central within the last 15 years and also gated private content from authoritative publishers such as Oxford University Press, Springer Nature, Wiley, Proceedings of the National Academy of Sciences (PNAS), the Journal of the American Medical Association (JAMA), Taylor & Francis, and more. 

Proprietary Data 

User may also integrate proprietary data into the ASCEND platform, enhancing the richness of their research workflows. For example, users can upload their Electronic Lab Notebooks (ELNs) to include experimental data, which captures workflows, observations, and results for historical experiments. Also, users can integrate ASCEND with their inventory systems and purchase history that track reagents, materials, and equipment used in experiments. This integration ensures that internal data is easily accessible alongside external sources, providing a comprehensive, streamlined view of the research process while maintaining the privacy and security of proprietary information.

🏢 Markets 

BioPharma R&D

Progress in drug development, measured by the number of new drugs approved each year, has been slowing steadily over the past 50 years—a phenomenon known as Eroom’s Law (Moore’s Law in reverse). Despite the explosion of technological innovation elsewhere, biopharma has struggled to translate scientific breakthroughs into new therapies at scale.

Meanwhile, R&D investment by major pharma companies has ballooned from $86B in 2013 (17.5% of sales) to $190B in 2023 (25.2% of sales), reflecting intensifying pressure to find the next wave of blockbuster drugs. This escalating investment, combined with stagnant productivity, creates an acute need for solutions that can drive efficiency and results. 

About 30-40% of the drug discovery cost is in preclinical R&D, which translates to over $60B spent at the early development stage for drug discovery, providing a large opportunity for BenchSci to compete. 

Drug Discovery Stage 

At the front end of the drug development pipeline, the challenge is even more pronounced. Biopharma companies today collect 7x more investigational drug data than they did two decades ago. Yet, only 5–6% of preclinical projects advance to clinical trials. Early-stage research remains heavily reliant on manual, siloed processes that are ill-suited for managing the complexity and volume of modern data.

AI in Life Sciences

The application of machine learning (ML) and natural language processing (NLP) techniques across the pharmaceutical value chain is still in its early innings. Most of the focus to date has been downstream—optimizing clinical trial management, patient recruitment, or drug commercialization. However, the largest unrealized value likely resides upstream, in preclinical R&D, where better hypothesis generation, experiment design, and risk identification could fundamentally alter the cost and success curves.

Companies can accrue benefits from AI in two ways: 

• Accelerate the Development Cycle 

• Achieve Cost Efficiency with the Development Process 

The current landscape for AI in Life Science is quite premature, however, two major waves are shaping the TechBio stack:

1. Data Infrastructure Evolution: → This is where BenchSci Plays
   Life sciences data is fragmented, similar to the early tech data stack. Before AI can have a real impact, the industry must integrate vast, multi-modal datasets (genomics, clinical trials, etc.) and modernize legacy platforms.
   

2. AI/ML Innovation Cycle:
   Advances like AlphaFold and Peptone highlight how tightly coupled data and AI are in life sciences. Rapid feedback loops between experiments and models are essential to improve predictions and accelerate drug discovery and development.

⚔️ Competitions

BenchSci is operating at the frontier of preclinical R&D enablement, but it is not alone. The competitive landscape can be divided into three broad categories:

AI Drug Discovery Platforms (Cyclica, Owkin, PostEra)

These firms apply AI to later stages of the drug discovery continuum (e.g., molecule generation, target-disease association). While their core competencies today differ (small molecule design, clinical prediction), the underlying technology trajectories are converging.

The real risk is platform creep. If players like Owkin (which is expanding into target identification) move upstream aggressively, BenchSci’s moat will rely heavily on maintaining and expanding its corpus advantage and embedding ASCEND deeper into pharma R&D operations.

Scientific Intelligence Platforms (Bioz, Zageno, Eagle Genomics)

These platforms provide AI-driven tools for reagent selection, procurement, or data integration, but typically focus on isolated stages of the research workflow. Many rely on citation aggregation or procurement simplification without deeply embedding into experimental hypothesis generation or validation. Others offer graph-based solutions but are often services-heavy and less scalable. 

Data Marketplaces (Antibodies.com, Antibodies-online, Biocompare, BIOZOL, IHC World)

These players aggregate reagent offerings but lack meaningful computational insight layers. They offer access to catalogues and data, but not experiment-specific recommendations. Switching costs remain low, and the decision-making burden is placed on the researcher. These sites may pose acquisition targets for downstream vertical integration, but do not currently challenge BenchSci’s strategic position.

⚙️ Business Model

BenchSci operates on a two-tiered SaaS model: one designed for academic users and one for enterprise pharmaceutical companies.

The academic tier allows researchers affiliated with educational institutions to access the platform, often with a limited set of features designed to accelerate experimental planning and reagent selection. This strategy helps BenchSci build early mindshare among future industry scientists while supporting open scientific innovation.

The enterprise tier targets large pharmaceutical, biotech, and life sciences companies. It provides a fully featured version of the ASCEND platform, including access to advanced applications like Selector, Navigator, Architect, and Defender, as well as enhanced AI assistants tailored to specific therapeutic areas and research stages. 

Pricing for enterprise clients is typically structured as an annual subscription with a seat-based fee structure. Additional fees based on data integrations and customizations may apply as well

💰Valuations & Fundraising

Since its inception, BenchSci has raised over CAD $218M across multiple funding rounds. In May 2023, BenchSci announced its Series D ($95M) led by Generation Investment Management with participation from existing investors Inovia Capital, TCV, Golden Ventures and F-Prime Capital.

The lead investors for previous rounds include: Real Venture (Pre-Seed), Golden Venture (Seed), iNovia Capital (Series A & Series C), Gradient Ventures (Series B), TCV (Series C). At the beginning, BenchSci had also participated in the accelerator program by Creative Destruction Lab.  

♟️ Key Opportunities

Growing Adoption of AI in Drug Discovery 

The biopharma industry is increasingly investing in AI-driven preclinical innovation, with 13.5% of R&D budgets allocated to this area in 2024, up from 9.8% in 2021. As pharma companies invest more into technology for early-stage drug discovery, particularly target identification and experimental design, BenchSci’s ASCEND platform is uniquely positioned to streamline the process and improve success probabilities for its customers.

With a growing track record of large BioPharma success with BenchSci’s platform, in addition to the growing appetite for AI capabilities, BenchSci can accelerate its GTM strategy and build real business cases for customers.

Growing Market Players in EBP 

Emerging Biopharma companies (EBPs) now account for a staggering 72% of the industry’s clinical pipeline (IQVIA, 2023). These firms are ramping up their R&D spend—$49B in 2023, a dramatic rise from $26B a decade ago—and face unique challenges as they lack the infrastructure of large pharma. As such, they are increasingly turning to external R&D enablement solutions like ASCEND, which provide scalable, turnkey applications (e.g., Selector, Navigator, Architect, Defender) that allow these companies to accelerate their experimental decision-making without the need for significant internal IT investment. BenchSci has a unique opportunity to capture a significant share of this market as EBPs continue to grow and diversify.

Expansion into Adjacent Vertical Markets

The ASCEND platform’s modularity positions it for expansion beyond its core preclinical R&D focus. There are clear opportunities for BenchSci to diversify into adjacent verticals, including:

• Computational Drug Design: Extending BenchSci’s AI capabilities into early-stage drug design to further accelerate time-to-market

• Companion Diagnostics: Providing critical insights into how biomarkers can guide personalized treatment options

• Biomanufacturing Optimization: Enhancing production processes to reduce costs and improve the scalability of biologics

• Translational Clinical Research: Leveraging preclinical insights to improve the transition to clinical trials, bridging the gap between bench and bedside

⚠️ Key Risks

Increasing Competition and Horizontal Platform Integration

BenchSci is competing in an increasingly crowded market. Companies like Recursion, Insitro, and Owkin are broadening their scope beyond preclinical experimental design to include full-stack AI-driven discovery platforms. If these players succeed in scaling horizontally, they could create greater competition against BenchSci's markets. Notably, Owkin raised $450M at a $ 1B+ valuation, signalling aggressive expansion into upstream R&D tools. If these competitors can scale across discovery, they may outpace BenchSci, reducing differentiation and limiting growth prospects.

Pharma’s Preference for Build over Buy 

The industry’s largest players are opting to build proprietary solutions rather than relying on third-party vendors. Novartis’ Nerve Live and Roche’s internal R&D capabilities are examples of how large pharma is consolidating data science and multi-omics research in-house. If pharma companies replicate BenchSci’s core functionalities internally, they may erode demand for external solutions like ASCEND, which could lead to downward pressure on pricing or the loss of key accounts. This “build vs. buy” dynamic is an increasing risk that could limit BenchSci’s long-term scalability if larger customers decide to invest heavily in internal capabilities.

Slow Adoption Cycle with Long Sales Cycle

Despite the compelling ROI BenchSci offers, pharma remains a notoriously slow adopter of new technologies, particularly in the early R&D stages. A Pistoia Alliance survey in 2023 revealed that only 28% of pharma professionals were actively adopting AI in early-stage R&D, citing concerns over validation standards and reproducibility. This slow adoption is further compounded by the long sales cycles (12-24 months), which require considerable upfront investment in customer education, integration, and ongoing support. BenchSci will need to continue investing heavily in building strong relationships and demonstrating the ROI of its platform before it sees meaningful adoption at scale.

Next week, we will be covering Stan, the Operating Platform for Creators and Digital Entrepreneurs

Resources 

• BenchSci Blog

• TechBio Market Map - B2Venture 

• Ascend Series D Announcement

• Tom Leung MSc PhD, Founder, BenchSci

• BenchSci Unveils Multimodel LLM’s Power to Revolutionize Perceptual AI