FAIR Workflows in Computational Materials Science
This course introduces a practical approach to organising multi-stage computational research as a coherent FAIR research object.
It shows how to move from storing individual files and final results towards a structured description of the entire research process: its stages, transitions, software environment, data provenance, researcher decisions, and links between intermediate and final results.
The course uses the MATSCI-NASU pilot package for the DFT → MLIP → MD workflow as a running example. For participants outside materials science, the same logic is presented through a generic research chain:
data collection → cleaning and selection → model training → model application → result analysis
MATSCI-NASU is presented as an example of a FAIR packaging approach for multi-stage computational research, not as a finished international standard.
Course Profile
Access: Open
Format: Online self-study course with presentations and full lecture texts
Language: Ukrainian
Structure: 4 thematic lectures
Duration: approximately 40 minutes per lecture
Practical module: Lecture 3 can be extended into an 80-minute workshop
Learning materials: 4 presentations + 4 full lecture texts
Level: Introductory / intermediate
Version: 1.0, 2026
Who Is This Course For?
The course is intended for:
- researchers and PhD students working with computational modelling, scientific data processing, or machine learning;
- researchers in materials science, physics, chemistry, and related disciplines;
- future data stewards and research data management specialists;
- data curators and repository staff;
- specialists working with research infrastructures and computing facilities.
Lectures 1, 2, and 4 do not require specialist knowledge of materials science.
Lecture 3 is designed for participants who want to move from understanding the concept to practical creation and validation of a FAIR research package.
What You Will Learn
By the end of the course, participants will be able to:
- explain why reproducibility is a property of the research process rather than of an individual file;
- distinguish between workflow specification, workflow execution, and provenance;
- decompose a multi-stage study into stages and transitions;
- explain the different roles of workflow, provenance, and manifest;
- identify the information required to trace a DFT → MLIP → MD workflow or a comparable research process;
- understand the roles of JSON Schema validation, package-level consistency checks, and expert scientific review;
- treat the software environment as part of the research result;
- work with controlled vocabularies, semantic crosswalks, and external identifiers;
- understand the path from a FAIR research package and DataverseUA towards OpenAIRE, EOSC, and other European research services;
- understand the strengths and limitations of automated FAIR assessment.
Running Example
MATSCI-NASU / SiC
The course uses a pilot computational workflow for silicon carbide as a running example:
DFT → training dataset → MLIP training → model selection/preparation → MD
The example is used to demonstrate:
- provenance of training data;
- transitions between computational stages;
- researcher decisions;
- model selection;
- the relationship between workflow, provenance, and manifest;
- construction of a FAIR research package.
For participants from other fields, the same logic can be mapped to:
data acquisition → cleaning/selection → model building → model application
Course Content
Lecture 1
From Open Data to Reproducible Processes: FAIR Workflows
Key idea
Reproducibility is a property of the process, not of an individual file.
A DOI assigned to a final result does not by itself make a study reproducible. To understand, verify, or repeat a computational study, we also need information about software versions, parameters, execution environment, intermediate objects, and the relationships between them.
Topics
FAIR principles · FAIR vs openness · workflow specification · execution · provenance · executable and descriptive workflows · international practices · where information is lost between research stages
Materials
Presentation
Full lecture text
Lecture 2
Modular Description of Multi-Stage Research: Model and Rationale
Key idea
A multi-stage study consists of stages and transitions, connected by three complementary descriptions: workflow, provenance, and manifest.
The lecture focuses especially on what happens at the boundaries between stages, where data are selected, transformed, filtered, or transferred and where human decisions are often made.
Topics
Stages and transitions · workflow as plan · provenance as execution history · manifest as package inventory · human decisions as provenance · 9 MATSCI-NASU modules · SiC pilot · comparison with NOMAD, AiiDA, and Workflow Run RO-Crate
Materials
Presentation
Full lecture text
Lecture 3
Practical Module: Metadata, Software Environment, and Validation
Key idea
A FAIR package requires three levels of control: schema validation, consistency checking, and expert scientific review.
JSON Schema checks the form.
Scripts check consistency between files.
A researcher or data steward checks whether the information is scientifically sufficient.
Topics
Schemas and examples · typical metadata errors · honest representation of unknown values · MODEL_CARD · manifest and SHA-256 · software environment · package consistency · schema-valid vs scientifically complete
Materials
Presentation
Full lecture text
Practical module
Some components of this module are already available, while the full hands-on environment is still being developed.
Lecture 4
Controlled Vocabularies, Crosswalks, EOSC Integration, and FAIR Assessment
Key idea
Interoperability begins with a shared machine-recognisable language.
The same scientific concept should be represented by the same or explicitly mapped terms and identifiers across different systems.
Topics
Controlled vocabularies · canonical values · semantic crosswalks · REUSE / PROFILE / MAP / LOCAL EXTENSION / LOCAL · external IRIs · PROV-O · DataCite · QUDT · ASMO · PMDco · JSON-LD · RO-Crate · DataverseUA → OpenAIRE → EOSC · automated FAIR assessment
Materials
Presentation
Full lecture text
Practical Module: Current Status
Already Available
- 9 JSON schemas and corresponding example records;
- MODEL_CARD generator;
- manifest generation and verification;
- DFT, MLIP, and MD metadata generators;
- guidelines for authors;
- current MATSCI-NASU draft package in DataverseUA.
Under Development
- Colab/Binder training environment;
- deliberately broken example records for validation exercises;
- one internally consistent educational mini-package;
- package-level consistency validator;
- extended exercises for researchers and data stewards.
This allows the course to be used now in an introductory format, while the full hands-on workshop can be added as the technical training environment matures.
Recommended Learning Sequence
Lecture 1 → Lecture 2 → Lecture 3 → Lecture 4
Lecture 1 explains why the research process itself needs to be described.
Lecture 2 shows how to structure a multi-stage study.
Lecture 3 turns the model into practical metadata and validation tasks.
Lecture 4 shows how to make the description understandable to other systems and connect it to the broader FAIR and EOSC ecosystem.
Course Materials
The course includes:
- 4 presentations;
- 4 full lecture texts;
- the MATSCI-NASU pilot package in DataverseUA;
- JSON schemas and example records;
- author guidelines;
- MODEL_CARD materials;
- semantic mapping materials;
- practical tools and exercises as they become available.
Open the MATSCI-NASU package in DataverseUA
Course Development
Course concept and structure: Natalia Denysiuk and the MATSCI-NASU team
Training materials: developed on the basis of the MATSCI-NASU work on FAIR description of multi-stage computational workflows in materials science
Pilot workflow: DFT → MLIP → MD for SiC
Educational and methodological support: Open Science and FAIR Data Competence Centre at Kyiv Academic University / Research Data Management Competence Centre of the National Academy of Sciences of Ukraine
Project context: MATSCI-NASU — preparatory work towards a thematic materials science node in the context of EOSC Gravity
Repository: DataverseUA
Course version: 1.0, 2026