Reproducibility in Science 2026

The RAS Dialogue Blog posts are written by RAS experts sharing the latest research, updates, and scientific RAS news. The content is curated by the RAS Initiative.

About the authors: Caroline DeHart, William Burgan and Carissa Grose are part of the RAS Initiative, Cancer Research Technology Program, Frederick National Laboratory for Cancer Research. Simona Colantonio is director of the Antibody Characterization Laboratory, Cancer Research Technology Program, Frederick National Laboratory for Cancer Research.

Image

Students, scientists, interdisciplinary experts, and other interested members of the international scientific community gathered on the Hood College campus (Frederick, MD) in June to tackle the complex challenges surrounding the improvement of reproducibility in science. 

Held in collaboration with the Frederick National Laboratory for Cancer Research, the "Reproducibility in Science 2026 – Elevating Standards to Strengthen Research" symposium featured internationally renowned subject matter experts from industry, academia, government, and scientific publishing, all of whom were passionate about improving scientific rigor, increasing public confidence in scientific research, and providing new perspectives on ways to improve other fields. Symposium topics, addressed in both thought-provoking plenary lectures and lively breakout panel discussion sessions, included antibody characterization, recombinant protein production, preclinical drug discovery, organoids and in vitro New Approach Methodologies (NAMs), genomic data from liquid biopsies, and AI scientific standards. 

Across these topics, the overarching theme was recognizing the importance of new reference standards and materials that would improve the replicability, reproducibility, and impact of each system during scientific research. 

Talks of interest to the RAS community

• Christopher Austin (GSK), who described the impact of unreliable reagents or assays on drug development and the moral imperative of maintaining the highest scientific standards for the benefit of future patients. 
• Joseph Wu (Stanford), who presented a workflow combining next-generation sequencing, CRISPR/Cas9, induced pluripotent stem cells (iPSCs), and AI-enhanced computational structure design to perform “clinical trials in a dish”, along with a validated bank of patient-derived iPSC lines freely available to academia and nonprofit organizations. 
• Jesse Boehm (Koch Institute, MIT), who described the importance of reproducible functional precision genomics in capturing the landscape of cancer diversity and dependencies, along with accelerating drug discovery through patient-driven tissue donation and updated DepMap tools, including KRAS inhibitor models. 
• Sarine Markossian (NCATS), who presented the NCATS Assay Guidance Manual Program and associated training opportunities for community-developed best practices in preclinical drug discovery and early translational research. NCATS hosts a collection of compound libraries for high-throughput screening assays and is currently developing a 1536-well nuisance compound library in collaboration with the Nuisance Compound Tool Consortium to serve as a high-quality reference tool to identify drug screening assay shortfalls and liabilities. 
• Jeffrey Whiteaker (Fred Hutch Cancer Center), who described the importance of proper biospecimen handling and validated proteomic assays run in a CLIA-CAP environment for reproducibly quantifying multiplexed targets, including RAS-MAPK pathway proteins, in heterogenous backgrounds such as patient tissue. 
• Aled Edwards (Univ. of Toronto), who presented the Structural Genomics Consortium Target 2035: Accelerating Chemical Probe Discovery Through Open Science program, which aims to help develop a pharmacological tool for every human protein. Target 2035 is seeking contributors and collaborators (soon to include the RAS Initiative) to generate high-quality protein-ligand screening datasets for open release to the scientific community and use in training predictive machine learning models, thus accelerating small molecule therapeutic development.

One of the hottest topics at the symposium was the importance of validating biological reagents, particularly antibodies. Mere days before, an independent collective of scientists had found that a major vendor may have manipulated the validation data for over 460 antibodies and that confusion between two similarly named antibodies had potentially impacted findings from over 300 publications. Antibodies play a vital role in biomedical research and the field has made meaningful progress, including stronger validation practices, Research Resource Identifiers (RRIDs) for traceability, stricter journal standards, community‑driven efforts like YCharOS and Only Good Antibodies, greater vendor transparency, and reduced lot‑to‑lot variability through recombinant technologies. 

However, antibodies still remain a major source of irreproducibility, leading to wasted resources, slower scientific progress, unnecessary use of human and animal samples, and erosion of public trust. Proper end-user education and rigorous antibody validation remain essential to prevent misapplication of otherwise high-quality reagents. As an example, the NCI Antibody Program’s highly characterized Kras4a (CPTC‑KRAS4a‑1) and HRas (CPTC‑HRAS‑1, ‑2, ‑3, ‑4) antibodies, available through the Developmental Studies Hybridoma Bank (DSHB), have been evaluated across multiple applications by the Antibody Characterization Lab (ACL) using established Standard Operating Procedures. These antibodies are valuable tools; however, researchers must still perform fit‑for‑purpose validation within their own experimental systems. This includes using appropriate positive and negative controls (e.g. as genetic knock-out or knock-down approaches) and ensuring that each antibody performs reliably in the specific technique and substrates relevant to their studies.

The symposium breakout panel discussions focused on the respective challenges and best practices for research conducted using antibodies or recombinant proteins, preclinical drug discovery, standardization of organoids or other NAMs, ensuring reproducibility in liquid biopsies, and developing scientific standards for AI. 

Shared threads throughout the symposium

Common themes reported by panel moderators included:

• The importance of comprehensive documentation 
• Reproducibility-focused experimental design 
• Good quality control 
• Rigorous validation 
• Standardized protocols 
• Hands-on training 
• Open data sharing 
• Data-driven methods 
• Replication studies 
• Community-derived best practices in elevating scientific reproducibility

Common challenges included:

• The complexity of sample populations and model systems 
• The resources and equipment available to a given laboratory 
• The time and labor costs required for validation
• The maturity gradient of technology and available reference standards 
• The need for context-dependent or “fit-for-purpose” reference materials 
•  The reality that several model systems (e.g., NAMs and AI) are too new for robust SOPs or validation methods to have been developed 

Panelists emphasized the need for sharing of negative data to be incentivized in order to streamline community knowledge and remove blind spots for the training of AI models, along with the importance of making raw data available via curated open-source repositories. Stakeholders, scientific journals, vendors, and end users were all exhorted to leverage their influence to ensure the highest standards for reagents, model systems, and SOPs, with several industry leaders held up as a positive example to follow. Each panel plans to publish white papers or manuscripts reporting on their collective findings, form new collaborations to fill standardization gaps, and make existing resources freely available to the scientific community.

While much work remains to be done, these combined efforts inspired by the symposium will directly benefit and accelerate new and ongoing research into human cancer and personalized medicine. Few things are impossible with commitment and enthusiasm, and we look forward to seeing the implementation of many of these new standards over the coming years.

Putting words into action

Scientific reproducibility is a core value for the RAS Initiative and the Frederick National Laboratory. Specifically, the sharing of a standardized set of reagents between the hub (Frederick National Laboratory for Cancer Research) and spokes (RAS investigators in government, academia, and industry) led the Protein Expression Laboratory, the provider of DNA, proteins, and RASless mouse embryonic fibroblast cell lines (generated from Dr. Mariano Barbacid’s DU1473 cell line [1]), to develop transparent, standardized processes for reproducible reagent generation. 

DNA constructs are deposited with Addgene, SOPs are freely accessible via the Frederick National Laboratory Scientific Standards Hub, and many detailed protocols have been published in peer-reviewed journals and Methods in Molecular Biology (e.g., [2], [3]). 

All RAS workflows, such as the production of a 15N labeled RAF1 (52-188) RBD-CRD protein using the Vibrio natriegens expression system [4], are performed by a large and dynamic team with a primary focus on optimizing yield while maintaining or exceeding the quality of the final protein. Quality control of RAS proteins is supported by the RAS Mass Spectrometry group through high-resolution intact mass and mass photometry analysis to assess the purity of recombinant proteins. A combination of top-down proteomics, targeted proteomics, and native mass spectrometry enables RAS Initiative scientists to elucidate the specific intricacies of the RAS post-translational modification landscape and gain insight into the complex stoichiometry and structural biology of RAS-MAPK proteins toward drug discovery efforts.

References

1. Matthias Drosten , Alma Dhawahir, Eleanor Y M Sum, Jelena Urosevic, Carmen G Lechuga, Luis M Esteban, Esther Castellano, Carmen Guerra, Eugenio Santos, Mariano Barbacid. Genetic analysis of Ras signalling pathways in cell proliferation, migration and survival. EMBO J. 2010 Mar 17;29(6):1091-104. doi: 10.1038/emboj.2010.7. 
2. Robert A D'Ippolito, Grace M Scheidemantle, Brian P Smith, Katie Powell^, Scott Eury, Abigail Neish, Jennifer Mehalko, Lauren Beaumont, Nicole Fer, Vanessa Wall, William Burgan, Anna E Maciag, Dominic Esposito, Caroline J DeHart. FLAG-KRAS4B as a Model System for KRAS4B Proteoform and PTM Evaluation by Mass Spectrometry. Methods Mol Biol. 2024:2797:299-322. doi: 10.1007/978-1-0716-3822-4_22.
3. Simon Messing, Constance Agamasu, Matt Drew, Caroline J DeHart, Andrew G Stephen, William K Gillette. Production and Membrane Binding of N-Terminally Acetylated, C-Terminally Farnesylated and Carboxymethylated KRAS4b. Methods Mol Biol. 2021:2262:105-116. doi: 10.1007/978-1-0716-1190-6_6.
4. Matthew Smith, José Sánchez Hernández, Simon Messing, Nitya Ramakrishnan, Brianna Higgins, Jennifer Mehalko, Shelley Perkins, Vanessa E Wall, Carissa Grose, Peter H Frank, Julia Cregger, Phuong Vi Le, Adam Johnson, Mukul Sherekar, Morgan Pagonis, Matthew R Drew, Min Hong, Stephanie R T Widmeyer, John-Paul Denson, Kelly Snead, Ivy Poon, Timothy Waybright, Allison Champagne, Dominic Esposito, Jane Jones, Troy Taylor, William Gillette. Producing recombinant proteins in Vibrio natriegens. Microb Cell Fact. 2024 Jul 24;23:208. doi: 10.1186/s12934-024-02455-5.

If you're interested in participating in the dialogue, consider joining RASLab. RAS Lab is an online discussion forum to facilitate the exchange of scientific information among researchers in cancers related to RAS. If you would like to join RAS Lab, send an email to SolveRAS@nih.gov with the subject line "I would like to join RAS Lab."