Scientific methods · Chapter 04.02

Molecular and pathway data

A molecular result becomes interpretable only when chemical identity, assay context, exposure and biological scale remain connected.

Methods, evidence and interpretation

Resolve the identity before joining the data

A gene symbol, protein accession and specific isoform are different identifiers. Use stable protein accessions, retain the organism and sequence version, and document identifier mapping. For compounds, preserve structure, stereochemistry, salt form and whether the record describes a parent compound or metabolite. Name matching alone can silently merge different entities; a technically successful database join does not guarantee biological equivalence.

[1][2]

Retain what an assay actually measured

A biochemical binding experiment, a cellular reporter assay and a viability endpoint answer different questions. Record assay format, target assignment, cell system, exposure duration and concentration units alongside each result. ChEMBL distinguishes assay categories and annotates target-assignment confidence. That confidence concerns the assay–target mapping; it is not a clinical evidence grade and must not be relabeled as a probability of patient benefit.

[2][5]

Do not treat IC50, Ki and Kd as interchangeable

Kd describes equilibrium dissociation in a specified binding system; Ki is an inhibition constant under a defined mechanism; IC50 is the concentration producing half-maximal inhibition in a particular assay. Under simple competitive enzyme inhibition, Ki = IC50/(1+[S]/Km). This conversion requires the corresponding kinetic assumptions and substrate information. Assay-dependent IC50 values cannot be universally converted into affinity, nor directly compared with an ingested amount.

[3]

Ask whether the experimental exposure is biologically relevant

Nominal concentration in a culture well is not necessarily free concentration at a human target. Protein binding, membrane passage, intracellular accumulation and metabolism can change exposure. An active metabolite may behave differently from its parent. Compare compatible concentration units and distinguish total from unbound measurements. Exposure plausibility constrains a mechanistic hypothesis; it does not by itself establish efficacy or safety.

[6]

Read pathway enrichment as a statistical question

Over-representation analysis asks whether more mapped entities fall in a pathway than expected under a specified background. The background should reflect the entities that could have been detected, not automatically every gene. Identifier coverage, overlapping pathways and multiple testing affect interpretation. A false discovery rate adjustment addresses a family of statistical tests; an enriched pathway is neither a proven causal mechanism nor a guaranteed therapeutic opportunity.

[4]

Separate pathway membership from functional direction

A pathway annotation connects knowledge; it does not establish that every component is active in every tissue. Expression, protein abundance, post-translational modification, localization and reaction flux describe different layers. A compound–target edge can be direct, inferred or context-dependent. Preserve the underlying experimental provenance and contradictory observations rather than collapsing them into one universal arrow. Transfer to cancer contexts requires the appropriate tissue and disease evidence.

[2][4]

Interactive methods lab

Synthetic educational examples · no patient data or compound recommendations

From overlap to an enrichment hypothesis

A fictional universe contains 200 measurable genes; 30 belong to one pathway. Change the size of a selected list and its overlap with that pathway.

Selected genes n20Pathway overlap k6

	In pathway	Outside pathway
Selected	6	14
Not selected	24	156

Expected overlap under the null: 3.00
One-sided hypergeometric p: 0.057
BH-adjusted p (10 tests): 0.095

P(X ≥ k) = Σ C(30,j)·C(170,n−j) / C(200,n).

The other nine p-values are fixed at 0.0005, 0.003, 0.009, 0.02, 0.04, 0.09, 0.2, 0.4 and 0.7. Benjamini–Hochberg adjustment is recalculated for the full family.

The universe and all values are invented. An appropriate measurable background, identifier mapping and a declared testing family are essential. Gene dependence and overlapping pathways affect interpretation; enrichment does not establish pathway activation or causality.

Methods behind the example: [7]

Scientific references

These sources support the scientific methods discussed. They do not validate Curetanics or establish efficacy for a compound. Reference titles remain in their original language.

← Back to homepage · Science & Evidence