BioDev & Data Curation | Bioinformatics, software, and data

Applied technical foundation

Biology, software, and data in one learning pathway

BioDev & Data Curation brings together three complementary areas of study and applied development: bioinformatics, software and web development, and data engineering.

The pathway begins with a biological question, then moves through source selection, data organization, analysis, and presentation. Documented prototypes make the methods reviewable and support continuity as the work develops.

Before continuing, review record organization, measurement scales, and biological-data formats.

01 / Biological information

Bioinformatics

Organizes biological entities, sequences, structures, functions, variants, and relationships so that scientific questions can be examined through traceable computational data.

Current emphasis: molecular identity, structural evidence, cannabinoid biosynthesis, receptor biology, and source curation.

02 / Digital implementation

Software and web development

Turns scientific requirements into maintainable code, accessible interfaces, visualizations, and modular integrations, with attention to security and data protection.

Current emphasis: semantic HTML, responsive CSS, JavaScript, scientific viewers, static publishing, and plans for application programming interfaces (APIs).

03 / Reliable data flows

Data engineering

Defines how distributed records are acquired, normalized, validated, stored, related, and delivered while preserving metadata, provenance, quality controls, and lineage.

Current emphasis: collecting and validating records, relational modeling, consistent identifiers, and preparing data for analysis and retrieval-augmented generation (RAG).

Knowledge area · AI

Artificial intelligence in research and data analysis

Artificial intelligence is used for classification, prediction, information retrieval, and analytical automation. In the Cannabis ecosystem, these applications include biological research, cultivation, regulation, specialized content, and information products. ^[1]

Its value depends not only on the model, but also on the quality of the data, the infrastructure that supports it, the people who develop and use it, and the governance that keeps its limits visible. ^[2]

Explore the AI knowledge area

Data quality is critical to useful AI models and can help reduce avoidable bias and unnecessary rework.

Data-oriented AI

Criteria for developing and evaluating AI applications

This perspective shifts attention from the isolated algorithm to the complete information lifecycle: origin, context, preparation, validation, interpretation, and use. It also includes the development of specialized content and information products whose sources, methods, review status, and intended use remain visible. AI-generated outputs are treated as working material until they have passed editorial and domain review.

OECD AI Principles

A shared framework for trustworthy artificial intelligence

The OECD principles connect innovation with human rights, transparency, safety, and accountability across the AI lifecycle.

01 / Inclusive growth Well-being and sustainable development
02 / Human values Rights, fairness, diversity, and privacy
03 / Transparency Understanding and challenging AI outputs
04 / Robustness Security, safety, and continuous risk management
05 / Accountability Traceability across the system lifecycle

International policy reference Read the OECD AI Principles Five principles for trustworthy AI across its lifecycle AromaCann · Explore AI and model governance

01 / Data quality

Documented, representative data reduce avoidable model errors

Metadata identify meaning and origin; consistency checks expose incompatible records; and validation criteria determine whether a dataset is suitable for the stated analysis.

02 / Operational efficiency

Versioned workflows make errors easier to locate and analyses easier to repeat

Recorded inputs, parameters, software versions, and review steps allow teams to reproduce an analysis, locate failures, and avoid repeating work whose prior conditions are unknown.

03 / Bias and limitations

Models require human review and generalization tests

Human evaluation, generalization tests, and documented uncertainty keep results within the scope that the data can actually support.

04 / Knowledge products

AI can support research, writing, and information retrieval

AI can support literature discovery, technical writing, educational resources, structured reports, dashboards, and retrieval interfaces when source records and human review remain part of the production chain.

Sources for the AI section 8 references · cited evidence and further reading

Traceable evidence

Evidence cited in the text

2 references

[1] OECD Policy instrument
Recommendation of the Council on Artificial Intelligence
OECD/LEGAL/0449 · adopted 2019, amended 2024.
[2] IMF Work and economy
Gen-AI: Artificial Intelligence and the Future of Work
Staff Discussion Note 2024/001 · January 2024.

Context by theme

Structural data, models, and molecular interpretation

Structural bioinformatics applies computational techniques, algorithms, and scientific software to the study of biological systems through biomolecular structure.

It brings together molecular biology, computing, physics, and chemistry to relate gene and protein sequences and annotations to experimental coordinates, predicted structures, molecular interactions, and functional information. These layers support three-dimensional models used to investigate molecular organization, interactions, and possible behavior.

The field extends from curated structural records to large-scale comparative analysis, simulation, data-intensive methods, and machine learning. In Kannabium, every interpretation must retain its source, molecular context, evidence class, analytical method, and uncertainty.

Identity: Organism, gene, transcript, protein sequence, isoform, experimental construct, variant, and persistent identifiers.
Structure: Experimental coordinates, predicted models, domains, motifs, residues, and missing regions, with experimental resolution or model-confidence measures as appropriate.
Molecular context: Substrates, products, cofactors, ligands, interaction partners, conformational states, membranes, and modifications.
Analysis: Comparison, visualization, modeling, simulation, data-intensive methods, machine learning, validation, and uncertainty.

Molecular-evidence criteria

Questions for evaluating molecular records

Use these questions to examine each molecular case. Every answer should identify its source and distinguish direct observation from annotation, prediction, simulation, and hypothesis.

01 Identity, gene, and sequence

Which organism, gene, locus, transcript, and annotation release define the biomolecule?: The answer must expose canonical identifiers, database sources, assembly or annotation context, and review status.
Which protein sequence is associated, and which molecular construct was studied?: The sequence accession, isoform, length, engineered changes, tags, mutations, and unresolved segments must remain distinguishable.
Which variants, paralogs, orthologs, or related sequences are relevant?: Relationships require explicit comparison criteria and must not be treated as equivalent function without supporting evidence.

02 Structure and functional features

Which experimental structures and predicted models are available, and what does each represent?: Method, resolution or confidence, construct, chain, state, ligands, missing regions, source record, and model version must accompany the visualization.
Which domains, motifs, residues, modifications, and molecular sites are functionally relevant?: Residue numbering and structural mapping must be traceable, with observed features separated from transferred or inferred annotations.

03 Mechanism, interactions, and molecular state

What molecular process is being examined: catalysis, ligand recognition, conformational change, or signaling?: A defensible answer connects structural observations to biochemical or biophysical evidence without presenting a static model as a complete mechanism.
How does THCA synthase convert cannabigerolic acid into tetrahydrocannabinolic acid, and what role does FAD play?: The enzyme module must connect reaction chemistry, cofactor state, catalytic residues, substrate interpretation, and the limits of the available experimental structure.
How do ligands, receptor states, signaling partners, and membrane context affect interpretation of CB1R?: The receptor module must compare constructs and conformations while separating binding poses, structural states, signaling evidence, and simulated behavior.

04 Evidence, prediction, and uncertainty

Which evidence supports each annotation?: Every claim should connect to a source record, publication, method, evidence class, review state, and curatorial note.
Which data are experimental, and which are predicted, modeled, transferred, or simulated?: These categories must remain visible in tables, visualizations, downloads, and explanatory text.
Which findings remain uncertain, incomplete, conflicting, or outside the validated scope?: Missing data, alternative interpretations, model limits, unresolved regions, and unsupported extrapolations must be disclosed rather than silently combined.

Applying the evaluation criteria

Two cases connect molecular records to their evidence limits

The architecture begins with one plant enzyme and one vertebrate receptor. Future biomolecules can enter the same framework without changing its evidence requirements.

Plantae · enzyme · biosynthesis

THCA synthase

Connects gene and protein identity to the experimental structure, catalytic residues, substrate context, FAD-dependent chemistry, variants, and the conversion of cannabigerolic acid into tetrahydrocannabinolic acid.

Examine THCA synthase evidence

Animalia · receptor · signaling

Cannabinoid receptor type 1

Connects CNR1 and its protein sequence to experimental receptor structures, ligands, conformational states, signaling partners, membrane context, comparative models, and interpretation limits.

Compare CB1R structural states

Machine learning foundation

How to prepare data and evaluate models

Machine learning comprises computational methods that use data to estimate patterns, classify observations, explore structure, or support decisions. A sound workflow connects a defined question, suitable data, and evaluation criteria to model choice and the interpretation of results.

In the Kannabium context, this means treating chemical profiles, biological observations, environmental records, market signals, and scientific documents as records whose origin and quality need to be known. Models require evaluation on suitable data and examination for bias, data leakage, overfitting, uncertainty, and limits of scope. Their outputs are not causal evidence by default.

01 / Learning paradigms

Training signals used by each learning paradigm

Supervised learning uses labeled examples for regression or classification. Unsupervised learning explores structure in unlabeled data. Reinforcement learning improves actions through interaction with an environment and reward signals. Deep learning uses multilayer neural networks and can be applied within different learning paradigms.

02 / Governed workflow

Records required for a reproducible and reusable result

Source and dataset with metadata and provenance.
Documented treatment, variables, and prediction target.
Algorithm, trained model, and partitioning strategy.
Validation, limitations, and intended application.

03 / Portal route

Applications to chemical, genomic, and environmental data

The machine-learning module introduces possible analyses of chemical profiles, spectroscopy, genomics, and environmental records. Its case study examines variable selection, validation, and interpretive limits.

Information products

Kannabium is a data-oriented information platform that integrates, processes, analyzes and visualizes scientific, technological, social, environmental and market data spanning plant biology, vertebrate receptor biology and the global bioeconomy of Cannabis sativa.

The portals, maps, and case studies let visitors consult sources, explore visualizations, and examine methods of analysis. BioDev develops the technical resources that support this presentation of knowledge.

Development follows criteria for data quality, provenance, metadata, traceability, interoperability, and scientific context when records are selected, transformed, and published.

Inspect platform capabilities and governance Technical architecture · provenance · responsible reuse

Areas of development

Integration of heterogeneous scientific and business data
Preparation and validation of datasets for analysis
Biological, technological and market data analysis
Environmental and social data integration
Interactive dashboards and scientific visualizations
Artificial intelligence-assisted data exploration
Metadata, provenance and traceability management
Production of sourced technical and educational content

Data governance foundation

AromaCann provides the data curation and governance foundation that supports Kannabium. This foundation defines data quality criteria, classification standards, metadata requirements, provenance controls and responsible data-use practices.

Open the AromaCann governance portal

BioDev & Data Curation applies these governance principles to technical architectures, analytical workflows, and research prototypes. It documents the methods behind Kannabium's published information resources.