Why Science? | SCENIC+, Enhancer-Driven Regulons and Single-Cell Multiomic Evidence

Section 1 of 36

1. Start with enhancer-driven regulation

SCENIC+ asks which transcription factors may act through accessible enhancer regions to influence candidate target genes in particular cell states. It integrates chromatin accessibility and expression, creating a richer chain than RNA alone while remaining an inference.

Contents · Next section

Section 2 of 36

2. Identify candidate enhancers

Chromatin-accessibility profiles reveal genomic regions open enough to be candidates for regulation. Accessibility is opportunity, not proof of enhancer function. Some open sites are inactive; some functional sites are difficult to detect.

Contents · Previous section · Next section

Section 3 of 36

3. Discover regulatory topics

The workflow uses pycisTopic to organise accessible regions into co-accessible topics and cell states. Topic models summarise structure but depend on feature filtering, topic number and sampling.

Contents · Previous section · Next section

Section 4 of 36

4. Find enriched motifs

pycisTarget tests whether transcription-factor motifs are enriched in candidate regions. Motif enrichment supports possible factor binding, but related factors can share motifs and sequence presence is not observed occupancy.

Contents · Previous section · Next section

Section 5 of 36

5. Link enhancers to genes

SCENIC+ associates candidate enhancers with target genes using accessibility and expression relationships, genomic information and filtering. Proximity alone is not enough because enhancers can act over distance and skip nearby genes.

Contents · Previous section · Next section

Section 6 of 36

6. Build enhancer-driven regulons

An eRegulon links a transcription factor, candidate enhancer regions and candidate target genes. Positive or negative relationships may be represented. Each three-part path should remain visible so the attractive TF name does not hide uncertain enhancer–gene links.

Contents · Previous section · Next section

Section 7 of 36

7. Keep the 2023 boundary

The Nature Methods study introduced SCENIC+ and demonstrated enhancer-driven GRN inference across studied multiomic settings. Current documentation describes beta software that may change. Cite the paper for evaluated claims and the pinned software for implementation.

Contents · Previous section · Next section

Section 8 of 36

8. Create a multiomic manifest

Record organism, tissue, condition, specimen, assay, cell or nucleus identifiers, genome build and whether RNA and ATAC were measured in the same cells. Paired and unpaired designs require different integration assumptions.

Contents · Previous section · Next section

Section 9 of 36

9. Check RNA quality

Inspect library size, detected genes, mitochondrial fraction, doublets and state composition. Expression noise affects TF and target relationships even when chromatin data are excellent.

Contents · Previous section · Next section

Section 10 of 36

10. Check ATAC quality

Report fragments, transcription-start-site enrichment, nucleosome signal, doublets and peak coverage. Sparse or low-quality accessibility can create unstable topics and motif results.

Contents · Previous section · Next section

Section 11 of 36

11. Preserve cell matching

For paired multiome data, retain the exact cell barcode relationship. For separate assays, document how cells or metacells were aligned. An apparent regulatory chain can arise from mismatched populations.

Contents · Previous section · Next section

Section 12 of 36

12. Pin genome resources

Save genome build, gene annotation, motif database, region rankings and species mappings. Coordinate mismatches can silently move or discard regulatory regions.

Contents · Previous section · Next section

Section 13 of 36

13. Choose topic settings

Record feature selection, number of topics, seeds and quality criteria. Compare defensible topic solutions and show whether leading eRegulons survive rather than depending on one decomposition.

Contents · Previous section · Next section

Section 14 of 36

14. Pin the software stack

Save SCENIC+, pycisTopic, pycisTarget and database versions, environment and command configuration. The official documentation warns that beta functions can change, making versioning especially important.

Contents · Previous section · Next section

Section 15 of 36

15. Practise with a fictional eRegulon table

This classroom table is invented and is not SCENIC+ output.

Contents · Previous section · Next section

Section 16 of 36

16. Read all three nodes

For every headline eRegulon, show factor expression, enhancer accessibility and target expression. A strong pattern in only one layer does not support the entire chain.

Contents · Previous section · Next section

Section 17 of 36

17. Inspect motif-family ambiguity

When several factors recognise similar motifs, use expression, footprints, cofactors or binding evidence to narrow identity. Otherwise report the motif family.

Contents · Previous section · Next section

Section 18 of 36

18. Inspect enhancer–gene distance

Plot genomic distance and compare distal with promoter-proximal links. Distance is informative but not a verdict. Chromatin-contact data can add orthogonal support.

Contents · Previous section · Next section

Section 19 of 36

19. Compare cell states

Show eRegulon activity across specimens and states, not only clusters pooled together. A state-specific programme should recur in independent biological samples.

Contents · Previous section · Next section

Section 20 of 36

20. Compare with SCENIC

Expression-only SCENIC and multiomic SCENIC+ answer related but different questions. Agreement can strengthen a regulon; disagreement can reveal missing accessibility, motif ambiguity or alignment problems.

Contents · Previous section · Next section

Section 21 of 36

21. Use perturbation simulation carefully

Current SCENIC+ documentation includes perturbation workflows, but simulated change remains model-based. Separate current software features from the 2023 paper and validate predictions experimentally.

Contents · Previous section · Next section

Section 22 of 36

22. Audit pairing assumptions

Unpaired RNA and ATAC data require alignment or metacells. Vary matching settings and show whether eRegulons persist. A perfect-looking chain built from mismatched states is not regulatory evidence.

Contents · Previous section · Next section

Section 23 of 36

23. Audit region sparsity

Single-cell ATAC is sparse. Aggregate appropriately, show coverage and test whether rare-state links survive resampling. Do not equate undetected accessibility with closed chromatin.

Contents · Previous section · Next section

Section 24 of 36

24. Audit motif databases

Motif collections differ in coverage and redundancy. Report the database and clustering rules, and test whether factor conclusions remain when similar motifs are grouped.

Contents · Previous section · Next section

Section 25 of 36

25. Audit enhancer–gene linking

Compare correlation, distance and any additional linking evidence. A link supported by one metric deserves weaker language than one with contact, perturbation or reporter evidence.

Contents · Previous section · Next section

Section 26 of 36

26. Audit composition and batch

RNA and ATAC batches can align to condition or cell type. Visualise modalities and specimens separately, use appropriate integration and retain genuine biological differences.

Contents · Previous section · Next section

Section 27 of 36

27. Design a decisive enhancer test

Choose a replicated TF–enhancer–gene chain. Perturb the factor or enhancer, measure accessibility, binding and target response, include nearby-region controls and test rescue or reporter activity.

Contents · Previous section · Next section

Section 28 of 36

28. Report an eRegulon evidence card

Include factor, motif family, enhancer coordinates, accessibility, target, link metric, cell-state recurrence, database versions, orthogonal evidence and perturbation status.

Contents · Previous section · Next section

Section 29 of 36

29. Connect to school science

SCENIC+ links DNA sequence, chromatin, transcription factors and gene expression in one evidence chain. Students can ask what each assay measures and where inference begins.

Contents · Previous section · Next section

Section 30 of 36

30. Build a three-card regulation lesson

Use factor, enhancer and gene cards. Let learners connect chains using motif, accessibility and expression clues, then reveal a perturbation result. The activity makes multiomic reasoning tangible.

Contents · Previous section · Next section

Section 31 of 36

31. Connect to computing

Topic models, motif databases, genomic intervals and multimodal joins show how software coordinates very different data types. Versioning and identifiers become scientific necessities.

Contents · Previous section · Next section

Section 32 of 36

32. Connect to mathematics

Matrix factorisation, enrichment, correlation and resampling provide quantitative links. Students can discuss why adding a modality can strengthen inference without guaranteeing causation.

Contents · Previous section · Next section

Section 33 of 36

33. Connect to career pathways

Enhancer analysis supports genomics, developmental biology, biotechnology and bioinformatics. Preparation spans biology, coding, statistics, data stewardship and experiments, without guaranteeing admission or work.

Contents · Previous section · Next section

Section 34 of 36

34. Use AI with guardrails

AI can explain eRegulons or help with plotting, but it may invent coordinates, motifs or package features. Verify against the paper, official repository and current documentation.

Contents · Previous section · Next section

Section 35 of 36

35. Run the publication checklist

Verify title, slug, paired-design claim, genome build, motif resources, software beta status, fictional table, sources, internal links and causal boundaries. Use ‘candidate enhancer’ and ‘predicted link’ precisely.

Contents · Previous section · Next section

Section 36 of 36

36. Finish with the right claim

SCENIC+ connects accessible regulatory DNA, transcription factors and gene responses into testable eRegulons. Its strength is a richer hypothesis chain and clearer perturbation target—not automatic proof of enhancer causality.

SCENIC+ results become clearer when every eRegulon is shown as a three-column chain: transcription factor, enhancer region and target gene. Add motif evidence, accessibility, enhancer–gene association and cell-state recurrence in adjacent columns. This prevents a complex multiomic model from collapsing into a single regulator label.

Did you know? Open chromatin does not mean an enhancer is actively driving a gene. Accessibility makes binding possible, while motif enrichment suggests a compatible factor and coordinated expression suggests a target relationship. The chain becomes stronger as these clues agree, yet perturbing the enhancer remains the cleanest route toward causal function.

Paired multiome designs measure RNA and chromatin accessibility in the same cell or nucleus, reducing alignment ambiguity. Separate scRNA-seq and scATAC-seq assays require computational matching or metacells. Report which design was used and show sensitivity to the matching rule. A beautiful chain assembled from mismatched states can be entirely misleading.

Genome coordinates need first-class provenance. Store genome build, gene annotation, region definitions and any liftover. A coordinate can appear valid while pointing to a different sequence assembly. For each leading enhancer, publish the exact interval and the evidence linking it to the target gene.

Motif-family ambiguity should be explicit. Closely related factors can recognise similar motifs, so the sequence evidence may support a family rather than one protein. Factor expression, footprints, cofactors and direct binding can narrow identity. Until then, family-level wording is more scientifically precise than an overconfident name.

Topic models and region filtering create analytical choices before eRegulons exist. Compare defensible topic numbers, random seeds and feature sets. Report whether leading enhancer–gene–factor chains recur. If a result depends on one decomposition, label it exploratory and preserve the alternative solutions.

Validation can be modular. CRISPR interference or deletion of the candidate enhancer tests necessity; reporter assays test sequence-dependent regulatory capacity; ChIP or CUT&Tag tests factor occupancy; RNA measurements test target response; rescue can strengthen specificity. Each assay answers one link, and together they can turn an eRegulon into a mechanism.

The current documentation warns that SCENIC+ remains under development and beta functions may change. Pin versions of SCENIC+, pycisTopic, pycisTarget and motif databases, save the environment and distinguish current tutorials from the 2023 publication. Software maturity is part of reproducibility, not a footnote.

For learners, this is a joyful example of scientific teamwork. Molecular biology creates the assays, computing aligns modalities, mathematics extracts patterns and experiments test function. No single view sees the whole regulatory system, but careful integration lets each view strengthen the others.

The right conclusion is hopeful and bounded: SCENIC+ can prioritise enhancer-driven regulatory programmes with unusually rich single-cell context. It cannot observe every molecular event. Its reward is a sharper target—an enhancer, factor, gene, state and experiment—that can reveal how regulation really works.

Build a modality-balance dashboard before network inference. For each cell state and specimen, display RNA cells, ATAC cells, paired cells where applicable, expression depth, accessible regions and quality metrics. A state rich in one modality and sparse in the other may generate a compelling but weakly supported eRegulon.

Enhancer–gene links should be compared with a distance-matched null. Nearby regions naturally have more opportunity to correlate with genes. Match candidate regions by distance, accessibility and variability, then ask whether selected links exceed this background. Orthogonal chromatin-contact information can provide an additional, separately labelled layer.

Negative eRegulon evidence must be phrased gently. Failure to identify a link may reflect sparse ATAC counts, an incomplete motif database, incorrect cell matching or regulation through a promoter or factor not represented. Write ‘not recovered under this workflow’ rather than ‘the enhancer is inactive’. The distinction matters for both science and teaching.

State specificity can be quantified without inventing a category. Report activity distributions across cells and specimen-level recurrence, compare against matched random region sets and test sensitivity to clustering resolution. If specificity disappears when clusters are slightly changed, present the programme as a gradient rather than a discrete state marker.

Perturbation simulations should remain visibly separate from measured multiome evidence. Label simulated vectors, state the starting eRegulons and show extrapolation regions. A future software feature should not be backdated into the 2023 paper. Pin current documentation and distinguish evaluation from exploration.

When comparing SCENIC+ with SCENIC, hold the expression data and cell labels constant where possible. Then differences can be attributed more clearly to enhancer and accessibility evidence rather than unrelated preprocessing. Agreement supports a core regulon; added SCENIC+ links reveal which regulatory DNA may carry it.

The final figure can use an evidence ladder: accessible region, motif enrichment, factor expression, enhancer–gene association, target expression, specimen recurrence, direct occupancy and perturbation. Colour only the rungs actually present. The ladder communicates progress without converting partial support into a completed mechanism.

Data governance matters in human multiome studies because genomic data can be sensitive. Follow consent, controlled-access and institutional requirements, report aggregation appropriately and avoid individual-level claims. Responsible curiosity includes protecting the people whose cells made the analysis possible.

For classroom discussion, ask learners why measuring two kinds of data can strengthen a hypothesis without guaranteeing it. The answer introduces independent evidence, shared bias and causal testing—ideas useful far beyond genomics, including AI outputs, news evaluation and everyday decision-making.

Create an eRegulon stability matrix across specimens, topic settings, motif-family groupings and enhancer-linking thresholds. Track complete TF–enhancer–gene chains rather than only factor names. A regulator can appear stable while the nominated enhancer and target change, which implies a broader programme but a fragile mechanism.

When a candidate enhancer lies near several genes, report all plausible targets and the evidence separating them. Distance, correlation, co-accessibility and contact data can disagree. A CRISPR perturbation that measures several neighbouring genes is often more informative than a reporter focused on only the favoured target.

Use a regional negative control: a nearby accessible region matched for activity but lacking the enriched motif or enhancer–gene association. Perturbing both helps show whether the candidate sequence has specific regulatory function rather than a generic local chromatin effect.

Multiomic integration should not erase modality disagreement. If accessibility suggests a factor while target expression does not respond, show the mismatch. It may indicate priming, delayed response, motif ambiguity or a failed link. These alternatives are richer science than forcing one consensus score.

The public-facing promise can stay simple: SCENIC+ helps researchers move from ‘this cell state looks different’ to ‘this factor may act through this piece of DNA to influence this gene.’ Every word—may, through, influence—marks a testable link. Careful language turns computational complexity into a clear experimental invitation.

Archive both modality-specific matrices and the integrated eRegulon objects. A future analyst should be able to reproduce each RNA, ATAC and linking panel independently before rerunning the full workflow. This staged design makes it possible to identify whether a changed conclusion came from one assay, the alignment or the network model.

End with a concrete perturbation plan rather than a grand network claim. Name the enhancer coordinates, factor, target, cell state, early readout, later phenotype and negative control. SCENIC+ earns its place when it makes this plan sharper and more efficient, even if the experiment ultimately revises the proposed chain.

Publish uncertainty at the level of the full chain. A stable factor with several interchangeable enhancers is different from a stable enhancer linked to uncertain targets. Chain-level recurrence shows which component is reliable and prevents a broad regulatory theme from being narrated as one exact molecular route.

Contents · Previous section · Continue to the Science Learning Hub