Wilder Ruiz

Bulk RNA sequencing (RNA-seq) measures which genes are active, and approximately how strongly, across a population of cells or a tissue sample.

RNA molecules reflect gene activity. Bulk RNA-seq counts RNA from many cells together and converts those measurements into expression profiles that can be compared between samples, treatments or biological conditions. 1 Wang, Z., Gerstein, M. and Snyder, M. (2009) 'RNA-Seq: a revolutionary tool for transcriptomics', Nature Reviews Genetics, 10, pp. 57-63. ↗ 2 Conesa, A. et al. (2016) 'A survey of best practices for RNA-seq data analysis', Genome Biology, 17, 13. ↗

R is a programming language designed especially for statistics, data analysis and scientific graphics.

R is widely used in bioinformatics because it supports specialised scientific software, including Bioconductor, an ecosystem of R packages for analysing biological and genomic data. 1 R Core Team, 'An Introduction to R'. R Foundation for Statistical Computing. ↗ 2 R Core Team, 'The R Language Definition'. R Foundation for Statistical Computing. ↗

Python is a general-purpose programming language widely used for scientific computing, automation and data processing.

In research software, Python can be used to clean and transform datasets, automate repeated tasks, connect analytical steps and build reusable scientific-computing components. 1 Python Software Foundation, 'The Python Tutorial'. Python Documentation. ↗ 2 Python Software Foundation, 'The Python Language Reference'. Python Documentation. ↗

edgeR is a Bioconductor software package for finding meaningful changes in gene-expression count data.

RNA-seq produces counts of sequencing reads associated with genes. edgeR statistically compares these counts between conditions while accounting for biological variability between replicate samples. 1 Robinson, M.D., McCarthy, D.J. and Smyth, G.K. (2010) 'edgeR: a Bioconductor package for differential expression analysis of digital gene expression data', Bioinformatics, 26(1), pp. 139-140. ↗ 2 Chen, Y. et al. (2025) 'edgeR v4: powerful differential analysis of sequencing data with expanded functionality and improved support for small counts and larger datasets', Nucleic Acids Research, 53(2), gkaf018. ↗

Linear Models for Microarray Data (limma), combined with voom, provides a statistical approach for comparing gene expression in RNA sequencing (RNA-seq) experiments.

RNA-seq measurements become more variable as expression levels change. Voom estimates this relationship and gives measurements statistical weights; limma then uses those weights to test for expression differences between experimental conditions. 1 Law, C.W., Chen, Y., Shi, W. and Smyth, G.K. (2014) 'voom: Precision weights unlock linear model analysis tools for RNA-seq read counts', Genome Biology, 15, R29. ↗ 2 Ritchie, M.E. et al. (2015) 'limma powers differential expression analyses for RNA-sequencing and microarray studies', Nucleic Acids Research, 43(7), e47. ↗

Principal Component Analysis (PCA) and Multidimensional Scaling (MDS) simplify complex datasets so similarities and differences between samples can be visualised.

Gene-expression datasets contain measurements for thousands of genes at once. PCA and MDS compress this complexity into a small number of visual dimensions, helping reveal sample groups, unusual samples and major biological or technical differences. 1 Jolliffe, I.T. and Cadima, J. (2016) 'Principal component analysis: a review and recent developments', Philosophical Transactions of the Royal Society A, 374, 20150202. ↗ 2 Ritchie, M.E. et al. (2015) 'limma powers differential expression analyses for RNA-sequencing and microarray studies', Nucleic Acids Research, 43(7), e47. ↗

Gene Set Variation Analysis (GSVA) and Gene Set Enrichment Analysis (GSEA) examine groups of biologically related genes instead of interpreting genes one at a time.

A gene set is a predefined group of genes linked by a biological process or signature. GSEA asks whether such genes accumulate toward one end of a ranked result, while GSVA estimates how strongly a gene set is represented within individual samples. 1 Subramanian, A. et al. (2005) 'Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles', Proceedings of the National Academy of Sciences, 102(43), pp. 15545-15550. ↗ 2 Hanzelmann, S., Castelo, R. and Guinney, J. (2013) 'GSVA: gene set variation analysis for microarray and RNA-seq data', BMC Bioinformatics, 14, 7. ↗

Single-sample Gene Set Enrichment Analysis (ssGSEA) estimates the activity of biological gene sets separately within each sample.

A gene set is a group of genes representing a pathway, process or biological signature. ssGSEA ranks genes within one sample and produces a score showing how strongly each gene set is represented in that sample. 1 Barbie, D.A. et al. (2009) 'Systematic RNA interference reveals that oncogenic KRAS-driven cancers require TBK1', Nature, 462, pp. 108-112. ↗ 2 Hanzelmann, S., Castelo, R. and Guinney, J. (2013) 'GSVA: gene set variation analysis for microarray and RNA-seq data', BMC Bioinformatics, 14, 7. ↗

The Molecular Signatures Database (MSigDB) is a curated library of gene sets representing biological processes, pathways and experimentally observed signatures.

A gene set is simply a defined list of biologically related genes. MSigDB supplies thousands of these lists for methods such as Gene Set Enrichment Analysis (GSEA), Gene Set Variation Analysis (GSVA) and single-sample GSEA (ssGSEA). 1 Liberzon, A. et al. (2011) 'Molecular signatures database (MSigDB) 3.0', Bioinformatics, 27(12), pp. 1739-1740. ↗ 2 Liberzon, A. et al. (2015) 'The Molecular Signatures Database hallmark gene set collection', Cell Systems, 1(6), pp. 417-425. ↗

Gene-identifier mapping translates between different naming systems used to refer to the same genes or transcripts.

For example, one dataset may use human-readable gene symbols while another uses Ensembl database identifiers. Mapping connects these systems and also deals with renamed, outdated or ambiguous identifiers. 1 Durinck, S., Spellman, P.T., Birney, E. and Huber, W. (2009) 'Mapping identifiers for the integration of genomic datasets with the R/Bioconductor package biomaRt', Nature Protocols, 4(8), pp. 1184-1191. ↗ 2 Gentleman, R.C. et al. (2004) 'Bioconductor: open software development for computational biology and bioinformatics', Genome Biology, 5, R80. ↗

Pathway analysis asks whether groups of genes belonging to the same biological process or signalling pathway change together.

Instead of interpreting a long list of individual genes, pathway analysis can summarise changes into processes such as inflammation, cell division, metabolism or cellular signalling. 1 Khatri, P., Sirota, M. and Butte, A.J. (2012) 'Ten years of pathway analysis: current approaches and outstanding challenges', PLoS Computational Biology, 8(2), e1002375. ↗ 2 Subramanian, A. et al. (2005) 'Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles', Proceedings of the National Academy of Sciences, 102(43), pp. 15545-15550. ↗

Toxicogenomics studies how potentially harmful exposures change gene activity and other molecular measurements.

For example, RNA-seq can reveal which biological pathways respond after cells are exposed to a drug, chemical or nanomaterial. These molecular patterns can help investigate mechanisms associated with toxicity. 1 Alexander-Dann, B. et al. (2018) 'Developments in toxicogenomics: understanding and predicting compound-induced toxicity from gene expression data', Molecular Omics, 14(4), pp. 218-236. ↗ 2 National Research Council (2007) 'Toxicogenomic Technologies'. In: Applications of Toxicogenomic Technologies to Predictive Toxicology and Risk Assessment. Washington, DC: National Academies Press. ↗

Cancer transcriptomics studies patterns of gene activity in tumours, cancer cells and experimental cancer models.

A transcriptome is the collection of RNA molecules produced by cells. Comparing tumour transcriptomes can reveal differences in biological pathways, tumour states and responses to experimental treatments. 1 Weinstein, J.N. et al. (2013) 'The Cancer Genome Atlas Pan-Cancer analysis project', Nature Genetics, 45, pp. 1113-1120. ↗ 2 Hoadley, K.A. et al. (2018) 'Cell-of-Origin Patterns Dominate the Molecular Classification of 10,000 Tumors from 33 Types of Cancer', Cell, 173(2), pp. 291-304.e6. ↗

JavaScript is a programming language commonly used to make websites and browser-based applications interactive.

It can respond to user actions, update information on screen and communicate with other software services. JavaScript can also run outside the browser to support server-side application logic. 1 Ecma International / TC39 (2026) 'ECMAScript Language Specification'. ↗ 2 WHATWG (2026) 'HTML Living Standard'. ↗

TypeScript extends JavaScript by allowing developers to describe more clearly what kind of data a program expects.

For example, a program can specify that a sample identifier should be text or that an expression value should be numeric. These checks can reveal mistakes during development before the software is run. 1 Microsoft (2026) 'The TypeScript Handbook'. TypeScript Documentation. ↗ 2 Microsoft (2026) 'Everyday Types'. TypeScript Documentation. ↗

Node.js allows JavaScript to run on a server or computer rather than only inside a web browser.

This makes JavaScript useful for backend work: receiving requests, reading files, communicating with databases and coordinating other services behind a web application. 1 OpenJS Foundation (2026) 'Node.js Documentation'. ↗ 2 OpenJS Foundation (2026) 'HTTP'. Node.js API Documentation. ↗

Express.js is a lightweight framework that helps organise web services built with Node.js.

A framework provides reusable structure for software. Express can decide what should happen when a browser requests a particular address, check incoming information and pass the request to the appropriate part of the application. 1 OpenJS Foundation and Express.js Contributors (2026) 'Routing'. Express.js Documentation. ↗ 2 OpenJS Foundation and Express.js Contributors (2026) 'Using middleware'. Express.js Documentation. ↗

Hypertext Preprocessor (PHP) is a programming language widely used to generate web pages and process requests on a web server.

Unlike browser-side code, server-side PHP runs before a page is sent to the visitor. It can assemble interfaces, validate submitted information and connect the website to databases or other services. 1 The PHP Group (2026) 'PHP Manual'. ↗ 2 The PHP Group (2026) 'Language Reference'. PHP Manual. ↗

Structured Query Language (SQL) is used to work with organised database records; MariaDB is a database system that stores and manages those records.

A relational database organises information into tables that can be linked through shared fields. In scientific software this can store users, datasets, metadata, workflow state and other structured records. 1 MariaDB Foundation / MariaDB plc (2026) 'SQL Structure'. MariaDB Server Documentation. ↗ 2 MariaDB Foundation / MariaDB plc (2026) 'SQL Statements'. MariaDB Server Documentation. ↗

Plumber allows scientific functions written in R to be requested by other software through an Application Programming Interface (API).

An API is a defined way for two software components to communicate. A browser interface can therefore request an R calculation from Plumber and receive the result without needing to run the R code directly in the browser. 1 Schloerke, B. et al. (2026) 'An API Generator for R: plumber'. ↗ 2 Schloerke, B. et al. (2026) 'Introduction'. plumber Documentation. ↗

Python with FastAPI provides a backend service for structured scientific processing and research-synthesis workflows.

FastAPI is a Python framework for building Application Programming Interfaces (APIs), which let different software components communicate. In ZAR AI Research City, this Python/FastAPI layer prepares retrieved PubMed evidence and supports the research-synthesis workflow behind the browser interface. 1 FastAPI Project (2026) 'FastAPI'. Official Documentation. ↗ 2 Python Software Foundation (2026) 'The Python Tutorial'. Python Documentation. ↗ 3 Ruiz, W. (2026) 'Zomniverse Project: Learn Mode'. Available at: Zomniverse, ZAR AI Research City (Accessed: 16 August 2026). ↗

An Application Programming Interface (API) provides a defined way for different software components to exchange information.

Representational State Transfer (REST) is a common approach for organising web APIs around identifiable resources and standard web requests. For example, a browser might request a dataset, submit an analysis or retrieve a result through separate addresses. 1 Fielding, R.T. (2000) 'Representational State Transfer (REST)'. In: Architectural Styles and the Design of Network-based Software Architectures. Doctoral dissertation, University of California, Irvine. ↗ 2 Fielding, R.T., Nottingham, M. and Reschke, J. (2022) 'HTTP Semantics', RFC 9110. Internet Engineering Task Force. ↗

HyperText Markup Language (HTML) defines the structure of a web page, while Cascading Style Sheets (CSS) control how that structure looks.

HTML can define elements such as headings, forms, tables and buttons. CSS controls layout, spacing, typography and responsive behaviour so the same scientific interface can work across different screen sizes. 1 WHATWG (2026) 'HTML Living Standard'. ↗ 2 World Wide Web Consortium (2025) 'CSS Snapshot 2025'. ↗

Full-stack scientific software development means building the connected parts of a research application, from the browser interface to server logic, scientific computation and stored data.

For a biologist, one action can pass through several software layers: the interface sends a request, a service runs R or Python, a database records state and the result returns to the browser. The "stack" is this set of connected layers. 1 FAIR for Research Software Working Group (2022) 'FAIR Principles for Research Software (FAIR4RS Principles), version 1.0'. Research Data Alliance. ↗ 2 OpenAPI Initiative (2025) 'OpenAPI Specification, version 3.2.0'. ↗

Modular frontend architecture divides browser code into smaller, focused parts instead of placing all interface behaviour in one large file.

A module may handle navigation, a tooltip or one workflow stage. Separating responsibilities makes individual parts easier to understand, replace and test without disturbing unrelated behaviour. 1 Ecma International / TC39 (2026) 'ECMAScript Language Specification: Scripts and Modules'. ↗ 2 WHATWG (2026) 'HTML Living Standard: Scripting'. ↗

Server-rendered progressive enhancement starts with useful HyperText Markup Language (HTML) produced by the server, then adds JavaScript where richer interaction is helpful.

The main content and navigation can exist before extra browser behaviour is added. "Progressive enhancement" means beginning with a working baseline and layering additional features on top. 1 The PHP Group (2026) 'PHP and HTML'. PHP Manual. ↗ 2 World Wide Web Consortium (W3C) (2015) 'Graceful degradation versus progressive enhancement'. ↗

Responsive interface engineering makes the same web application adapt to different screen sizes and display conditions.

Cascading Style Sheets (CSS) media queries can change layout when available space changes, while accessible reflow keeps content usable on smaller screens without unnecessary two-direction scrolling. 1 World Wide Web Consortium (W3C) (2026) 'Media Queries Level 5'. ↗ 2 World Wide Web Consortium (W3C) (2024) 'Understanding Success Criterion 1.4.10: Reflow'. WCAG 2.2. ↗

Component-oriented Cascading Style Sheets (CSS) architecture keeps styling rules organised around individual interface parts such as cards, navigation controls or analysis panels.

Clear naming and style scoping reduce the chance that a rule intended for one component unexpectedly changes another, making larger interfaces easier to maintain. 1 World Wide Web Consortium (W3C) (2026) 'CSS Scoping Module Level 1'. ↗ 2 World Wide Web Consortium (W3C) (2025) 'CSS Snapshot 2025'. ↗

Research platform architecture defines how scientific interfaces, data, computational services and stored results are separated and connected.

Clear boundaries help a platform send the right data to the right analytical service while keeping software responsibilities, metadata and results traceable. This supports reuse and interoperability between research-software components. 1 FAIR for Research Software Working Group (2022) 'FAIR Principles for Research Software (FAIR4RS Principles), version 1.0'. Research Data Alliance. ↗ 2 OpenAPI Initiative (2025) 'OpenAPI Specification, version 3.2.0'. ↗

A database-backed application stores important information in a database rather than keeping it only in temporary browser memory or separate files.

In scientific software, a relational database can preserve users, datasets, metadata, workflow state and results as structured records that can be retrieved and linked later. 1 MariaDB (2025) 'Introduction to Relational Databases'. MariaDB Documentation. ↗ 2 MariaDB (2026) 'Relational Databases: Basic Terms'. MariaDB Documentation. ↗

A data-driven User Interface (UI) builds visible content or controls from structured data instead of repeating the same information directly in page code.

For example, one JavaScript Object Notation (JSON) record can supply a label, status, version and description that the interface renders consistently in several places. JSON is a standard text format for structured data. 1 Bray, T. (2017) 'The JavaScript Object Notation (JSON) Data Interchange Format', RFC 8259. Internet Engineering Task Force. ↗ 2 WHATWG (2026) 'HTML Living Standard'. ↗

Accessibility-aware interface design considers whether people with different abilities can perceive, understand and operate a web application.

This includes meaningful structure, keyboard access, visible focus and correctly described controls. Accessible Rich Internet Applications (ARIA) adds accessibility information when native HTML alone is not sufficient. 1 World Wide Web Consortium (W3C) (2024) 'Web Content Accessibility Guidelines (WCAG) 2.2'. ↗ 2 World Wide Web Consortium (W3C) (2026) 'ARIA Authoring Practices Guide (APG)'. ↗

Technical Search Engine Optimization (SEO) helps search engines discover and understand a website; structured data adds machine-readable meaning to page content.

Structured data can identify items such as a person, organisation or dataset in a standard form. It complements clear page titles, links, crawlable content and other technical signals used by search engines. 1 Google Search Central (2026) 'Search Engine Optimization (SEO) Starter Guide'. ↗ 2 Google Search Central (2026) 'Introduction to structured data markup in Google Search'. ↗

Deployment validation checks a software release before it replaces the live version of an application.

Checks can confirm required files, versions and configuration, while a "dry run" previews transfers or deletions without changing the live site. Release management records which software version was promoted and what changed. 1 Preston-Werner, T. (2026) 'Semantic Versioning 2.0.0'. ↗ 2 Rsync Project (2026) 'rsync(1) manual page'. ↗

Workflow orchestration coordinates the order in which different computational or analytical steps are performed.

For example, an RNA-seq workflow may need to validate a dataset before normalisation and complete quality control before downstream analysis. Orchestration manages these hand-offs and ensures each stage receives the information produced by the previous one. 1 Crusoe, M.R. et al. (2022) 'Methods Included: Standardizing Computational Reuse and Portability with the Common Workflow Language', Communications of the ACM, 65(6), pp. 54-63. ↗ 2 Common Workflow Language Working Group (2026) 'Common Workflow Language Workflow Description Specification, v1.2'. ↗

Browser-based scientific User Experience (UX) design makes computational research tools understandable and usable through a web browser.

Good scientific UX helps researchers understand what a tool needs, what stage an analysis has reached and what its outputs mean. It also considers accessibility, keyboard use, clear feedback and different screen sizes. 1 World Wide Web Consortium (2024) 'Web Content Accessibility Guidelines (WCAG) 2.2'. ↗ 2 World Wide Web Consortium (2026) 'ARIA Authoring Practices Guide (APG)'. ↗

Metadata-driven workflows use structured information about data to control how that data is processed.

Metadata simply means data that describes other data, such as sample identity, processing stage, source or version. A workflow can use those fields to decide what happens next instead of relying only on filenames or manual choices. 1 DataCite Metadata Working Group (2026) 'DataCite Metadata Schema Documentation for the Publication and Citation of Research Data and Other Research Outputs', Version 4.7. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV Model Primer'. ↗

Provenance records where data or an analytical result came from and how it was produced.

For research data, provenance can connect an output to its source data, processing steps and responsible activities, making the route from input to result easier to inspect. 1 World Wide Web Consortium (W3C) (2013) 'PROV-O: The PROV Ontology'. W3C Recommendation. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV Model Primer'. ↗

Secure Hash Algorithm 256-bit (SHA-256) produces a compact digital fingerprint of data.

A checksum is a value calculated from a file or message. Recalculating its SHA-256 value later can help detect whether the underlying data has changed. 1 National Institute of Standards and Technology (NIST) (2015) 'Secure Hash Standard (SHS)', FIPS PUB 180-4. ↗ 2 National Institute of Standards and Technology (NIST) 'Secure Hash Algorithm 256'. Computer Security Resource Center Glossary. ↗

Audit events are timestamped records of important actions or changes within a system.

An audit trail can record events such as a dataset action, state change or processing operation so that important system activity can later be reviewed. 1 Kent, K. and Souppaya, M. (2006) 'Guide to Computer Security Log Management', NIST Special Publication 800-92. ↗ 2 National Institute of Standards and Technology (NIST) (2020) 'Security and Privacy Controls for Information Systems and Organizations', SP 800-53 Rev. 5. ↗

A governed artifact is a saved research output managed with explicit rules about its identity, state and origin.

Here, an artifact can be a processed dataset, result file or other analytical output. Governance means the system keeps structured information around that object rather than treating it as an anonymous file. 1 RO-Crate Community (2026) 'RO-Crate Specification'. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV Model Primer'. ↗

An artifact lifecycle describes the stages a research output passes through from creation to later use or replacement.

An analytical artifact may be created, reviewed, accepted, superseded or reused. Recording those stages helps distinguish temporary outputs from results that have been intentionally retained. 1 RO-Crate Community (2026) 'RO-Crate Specification'. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV-O: The PROV Ontology'. W3C Recommendation. ↗

Authentication verifies who a user is; Cross-Site Request Forgery (CSRF) protection helps stop another site from submitting unwanted actions through that user's authenticated browser.

These controls address different problems: authentication establishes user identity, while CSRF protection helps ensure that a sensitive request genuinely originated from the intended application interaction. 1 Open Worldwide Application Security Project (OWASP) 'Authentication Cheat Sheet'. ↗ 2 Open Worldwide Application Security Project (OWASP) 'Cross-Site Request Forgery Prevention Cheat Sheet'. ↗

Owner-scoped datasets are research datasets whose access is restricted according to the authenticated user that owns or controls them.

The scope acts as an access boundary: a request should be authorised for the relevant dataset rather than assuming that authentication alone permits access to every stored dataset. 1 Open Worldwide Application Security Project (OWASP) 'Authorization Cheat Sheet'. ↗ 2 National Institute of Standards and Technology (NIST) (2020) 'Security and Privacy Controls for Information Systems and Organizations', SP 800-53 Rev. 5. ↗

Owner-scoped storage separates stored resources according to the user or identity authorised to control them.

The idea is similar to labelled laboratory storage: knowing that a file exists is not enough to grant access. The application also checks whether the requesting identity is permitted to reach that stored resource. 1 Open Worldwide Application Security Project (OWASP) 'Authorization Cheat Sheet'. ↗ 2 National Institute of Standards and Technology (NIST) (2020) 'Security and Privacy Controls for Information Systems and Organizations', SP 800-53 Rev. 5. ↗

Review and commit states separate a provisional result from one that has been deliberately accepted into the workflow.

A result can first remain inspectable as a candidate. Committing it records the decision to treat that state as the accepted output for the next stage, rather than silently overwriting earlier work. 1 Git Project (2026) 'git-commit: Record changes to the repository'. Git Documentation. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV Model Primer'. ↗

Structured issue codes are stable machine-readable identifiers for known warnings, errors or workflow conditions.

Instead of returning only free-text messages, software can attach a predictable code to a problem. The interface can then route that code to consistent guidance while still showing a human-readable explanation. 1 Nottingham, M., Wilde, E. and Dalal, S. (2023) 'Problem Details for HTTP APIs', RFC 9457. ↗ 2 Kent, K. and Souppaya, M. (2006) 'Guide to Computer Security Log Management', NIST Special Publication 800-92. ↗

Mistral is a family of large language models (LLMs); inference means running a trained model to generate an output from a new input.

Local Mistral inference means the model is served on controlled infrastructure rather than relying entirely on a remote hosted service. The LLM can then support explanation or language tasks while the surrounding application controls what information and tools it receives. 1 Mistral AI (2026) 'Inference'. Mistral AI Documentation. ↗ 2 Mistral AI (2026) 'Self-Deployment'. Mistral AI Documentation. ↗

Deterministic routing sends a request to a predefined tool or pathway using explicit rules rather than asking an artificial intelligence (AI) model to choose freely.

Here, routing simply means deciding which component should handle a request. Deterministic means that the same defined conditions lead to the same route, helping keep calculations, retrieval and generated explanation in separate roles. 1 Mistral AI (2026) 'Inference'. Mistral AI Documentation. ↗ 2 National Institute of Standards and Technology (NIST) (2023) 'Artificial Intelligence Risk Management Framework (AI RMF 1.0)', NIST AI 100-1. ↗

Structured retrieval means deliberately fetching relevant records, documents or evidence from an organised source before producing an answer or explanation.

Retrieval separates information lookup from language generation. Instead of relying only on information stored inside a large language model (LLM), the system can supply selected external material for the current request. 1 Lewis, P. et al. (2020) 'Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks', Advances in Neural Information Processing Systems, 33. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV-DM: The PROV Data Model'. ↗

Structured issue guidance links a known software or workflow problem to predefined explanatory information and possible next steps.

An issue can first be represented by a stable code or known condition. The interface can then retrieve consistent guidance for that condition rather than depending entirely on a large language model (LLM) to invent a diagnosis from free text. 1 Nottingham, M., Wilde, E. and Dalal, S. (2023) 'Problem Details for HTTP APIs', RFC 9457. ↗ 2 National Institute of Standards and Technology (NIST) (2024) 'Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile', NIST AI 600-1. ↗

Evidence-aware explanation means generated text is produced with explicit supporting information available to the system.

The aim is to keep the explanation connected to retrieved records or other identifiable evidence rather than treating generated language itself as the evidence. Provenance describes where information came from and how it was produced. 1 Lewis, P. et al. (2020) 'Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks', Advances in Neural Information Processing Systems, 33. ↗ 2 World Wide Web Consortium (W3C) (2013) 'PROV-O: The PROV Ontology'. W3C Recommendation. ↗

Bounded scientific assistance restricts artificial intelligence (AI) to defined supporting tasks rather than allowing it to become the sole authority for scientific computation or evidence.

Here, bounded means deliberately limited. AI can help retrieve, organise or explain information while deterministic calculations, recorded evidence and human scientific judgement remain distinguishable from generated language. 1 National Institute of Standards and Technology (NIST) (2023) 'Artificial Intelligence Risk Management Framework (AI RMF 1.0)', NIST AI 100-1. ↗ 2 National Institute of Standards and Technology (NIST) (2024) 'Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile', NIST AI 600-1. ↗

Gas Chromatography–Mass Spectrometry (GC-MS) combines chemical separation with mass-based identification and measurement.

Gas chromatography separates volatile compounds in a sample. Mass spectrometry (MS) then detects characteristic ions that can help identify or quantify the separated compounds. 1 Lovestead, T.M. and Urness, K. (2019) 'Gas Chromatography - Mass Spectrometry (GC-MS)'. National Institute of Standards and Technology. ↗ 2 National Institute of Standards and Technology (NIST) (2024) 'Mass Spectrometry Instrument Lab'. ↗

Liquid Chromatography–Tandem Mass Spectrometry (LC-MS/MS) combines liquid-phase chemical separation with two stages of mass spectrometry.

Liquid chromatography separates compounds in a sample before tandem mass spectrometry (MS/MS) selects and fragments ions for sensitive and selective measurement of target molecules. 1 U.S. Food and Drug Administration (FDA) (2018) 'Bioanalytical Method Validation: Guidance for Industry'. ↗ 2 International Council for Harmonisation (ICH) (2022) 'M10 Bioanalytical Method Validation and Study Sample Analysis'. ↗

A Laboratory Information Management System (LIMS) is software used to organise laboratory samples, workflows and associated data.

A LIMS can help laboratories track specimens, experimental records and data across stages of work, reducing reliance on disconnected spreadsheets or manual records. 1 National Institute of Standards and Technology (NIST) (2025) 'NIST Research Data Framework (RDaF)'. ↗ 2 Lau, J.W. et al. (2021) 'A Roadmap for LIMS at NIST Material Measurement Laboratory'. National Institute of Standards and Technology. ↗

Biological sample preparation converts material such as blood, plasma, cells or tissue into a form suitable for measurement or analysis.

Preparation can include steps such as extraction, dilution, separation or cleanup. Careful preparation is important because unwanted sample components can interfere with analytical measurements. 1 International Council for Harmonisation (ICH) (2022) 'M10 Bioanalytical Method Validation and Study Sample Analysis'. ↗ 2 U.S. Food and Drug Administration (FDA) (2018) 'Bioanalytical Method Validation: Guidance for Industry'. ↗

Quality control (QC) uses defined checks and reference samples to determine whether an analytical process is performing acceptably.

For example, a QC sample can contain a known amount of the substance being measured. Its result helps show whether an analytical run remains sufficiently accurate and consistent. 1 U.S. Food and Drug Administration (FDA) (2018) 'Bioanalytical Method Validation: Guidance for Industry'. ↗ 2 International Council for Harmonisation (ICH) (2022) 'M10 Bioanalytical Method Validation and Study Sample Analysis'. ↗

Pharmacology studies how drugs interact with biological systems and how those interactions produce desired or unwanted effects.

It includes questions such as what a drug does to the body and how the body absorbs, distributes, metabolises and eliminates the drug. These relationships help connect dose, exposure and biological response. 1 U.S. Food and Drug Administration (FDA) (2023) 'Clinical Pharmacology: Early Drug Development'. ↗ 2 U.S. Food and Drug Administration (FDA) (2024) 'Clinical Pharmacology Guidances: Advancing Drug Development and Regulatory Assessment'. ↗

Regulated laboratory workflows follow documented quality requirements so that testing is controlled, traceable and capable of producing reliable results.

Examples include defined procedures for equipment, samples, records, quality checks and reporting. Good Laboratory Practice (GLP) and ISO/IEC 17025 are examples of recognised laboratory-quality frameworks. 1 International Organization for Standardization (ISO) (2017) 'ISO/IEC 17025:2017 — General requirements for the competence of testing and calibration laboratories'. ↗ 2 Organisation for Economic Co-operation and Development (OECD) 'OECD Principles on Good Laboratory Practice'. ↗

Patient-derived cancer models are experimental tumour models established directly from material collected from people with cancer.

Examples include patient-derived tumour cell cultures, organoids and xenografts. Because they originate from patient material, these models can preserve clinically relevant biological features for experimental drug and cancer research. 1 National Cancer Institute (NCI) (2026) 'NCI Patient-Derived Models Repository (PDMR)'. ↗ 2 National Cancer Institute (NCI) (2026) 'PDMR Models'. ↗

Nanoparticle safety examines whether nanoscale materials can produce harmful biological or environmental effects and under what exposure conditions.

Nanomaterials can behave differently from conventional forms of a material, so safety assessment may need to consider properties such as particle size, aggregation, dispersion and delivered dose. 1 Organisation for Economic Co-operation and Development (OECD) (2026) 'NANOMET: Towards tailored safety testing methods for nanomaterials'. ↗ 2 Organisation for Economic Co-operation and Development (OECD) (2025) 'Guidance on Sample Preparation and Dosimetry for Manufactured Nanomaterials'. ↗