Source study found

Story checked

Scalable machine learning model maps complex genomic enhancers across tissues (opens in a new tab)

news-medical.net · 2026-09-17

Short answerEvidenceSource

Short answer

Mixed

Mixed.

One claim goes further than the study. 2 other points were not covered by the paper.

2 supported
1 overstated
2 not covered

Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.

Share this check

Follow the evidence trail

The story

Scalable machine learning model maps complex genomic enhancers across tissues

news-medical.net · 2026-09-17

The story’s checkable claims.

Read the original story

NewsLink checks it

Mixed

One claim overstates the study. Two of five check out. Two claims the study doesn't address.

2 supported
1 overstated
2 not covered

Open claim evidence

The source study

Mapping enhancer–gene regulatory interactions from single-cell data

Nature Genetics · 2026

Open source record Open in Tessa

Then inspect each claim

Evidence layer

Claim by claim

Each claim gets a verdict. Expand it to see the evidence directly below.

Reading mode

Scan verdicts. Open evidence only when needed.

Browse by verdict

5 claims in this story

Showing all 5 claimsChoose a verdict to focus the list.

Claim 1 of 5OverstatedThe researchers used the models in complex tissues to trace disease-associated variants to target genes, including linking INPP4B and IL15 to the number of lymphocytes in blood.View evidenceHide evidence

Why this verdict

The profile supports that scE2G was applied in heterogeneous tissues to interpret noncoding complex-trait variants and nominate regulatory interactions linking INPP4B and IL15 to lymphocyte count. The story’s phrasing that the models 'trace disease-associated variants to target genes' and 'link' these genes to blood lymphocyte number is stronger than the abstract evidence, which describes computational nomination and does not report experimental validation or quantitative support for the specific locus-level links.

Study evidence

Application of scE2G to single-cell chromatin/multiome data produced enhancer–gene regulatory interaction maps in heterogeneous tissues and was used to interpret noncoding complex-trait variants.

“We apply scE2G to build maps of enhancer–gene regulatory interactions in heterogeneous tissues and interpret noncoding variants associated with complex traits, nominating regulatory interactions linking INPP4B and IL15 to lymphocyte count.”

Claim 2 of 5Not coveredOnce trained, the models can be applied to data from other cell types and can build maps for rare or hard-to-isolate cell types using single-cell data.View evidenceHide evidence

Why this verdict

The abstract profile supports some generalization: scE2G is benchmarked across several cell types and applied in heterogeneous tissues. However, the specific claim that the trained models can build maps for rare or hard-to-isolate cell types is not present in the abstract-level evidence. That may be discussed in the full paper, but it is not verifiable at the supplied depth.

Study evidence

When benchmarked against CRISPR perturbations, fine-mapped eQTLs, and GWAS variant–gene associations, scE2G models are reported to achieve state-of-the-art performance for predicting enhancer–gene regulatory interactions across multiple cell types and perturbation categories.

“We benchmark scE2G models against CRISPR perturbations, fine-mapped expression quantitative trait loci and genome-wide association study variant–gene associations and demonstrate state-of-the-art performance at prediction tasks across several cell types and categories of perturbations.”

Study evidence

Application of scE2G to single-cell chromatin/multiome data produced enhancer–gene regulatory interaction maps in heterogeneous tissues and was used to interpret noncoding complex-trait variants.

“We apply scE2G to build maps of enhancer–gene regulatory interactions in heterogeneous tissues and interpret noncoding variants associated with complex traits, nominating regulatory interactions linking INPP4B and IL15 to lymphocyte count.”

Claim 3 of 5Not coveredThe team says the models perform well on datasets of varying sizes and sequencing depths, making them usable across many existing single-cell datasets.View evidenceHide evidence

Why this verdict

The paper profile says scE2G achieved state-of-the-art performance across several cell types and perturbation categories, but it does not report performance by dataset size or sequencing depth, nor does it establish usability across many existing single-cell datasets. Because the supplied evidence is abstract-level and lacks those benchmarking details, this claim is not verifiable at this depth.

Study evidence

“We benchmark scE2G models against CRISPR perturbations, fine-mapped expression quantitative trait loci and genome-wide association study variant–gene associations and demonstrate state-of-the-art performance at prediction tasks across several cell types and categories of perturbations.”

Claim 4 of 5SupportedA recent Nature Genetics study from Stanford University developed single-cell enhancer-to-gene prediction models, called scE2G, that predict genome-wide enhancer interactions from scATAC or multiomic scATAC and scRNA-seq data.View evidenceHide evidence

As statedgenome-wide

Why this verdict

The paper profile supports the core scientific claim: scE2G is introduced as a family of classification models to predict enhancer–gene regulatory interactions using features from single-cell ATAC-seq or multiomic RNA+ATAC data. At abstract depth, the profile does not independently substantiate the Stanford affiliation/provenance or give detailed implementation scope for 'genome-wide,' but the main model-development claim is aligned with the abstract.

Study evidence

The authors introduce scE2G, a family of classification models to predict enhancer–gene regulatory interactions from single‑cell data.

“Here we introduce a family of classification models, scE2G, to predict enhancer–gene regulation.”

Claim 5 of 5SupportedThe researchers trained the models using CRISPR experiments that directly tested more than 10,000 candidate enhancer–gene pairs.View evidenceHide evidence

As statedmore than 10,000 candidate enhancer–gene pairs

Why this verdict

The abstract-level profile states that scE2G models are trained on a CRISPR perturbation dataset including more than 10,000 evaluated element–gene pairs. This matches the story’s statement that CRISPR experiments directly tested more than 10,000 candidate enhancer–gene pairs, with only minor wording differences between 'evaluated element–gene pairs' and 'candidate enhancer–gene pairs.'

Study evidence

The authors introduce scE2G, a family of classification models to predict enhancer–gene regulatory interactions from single‑cell data.

“Here we introduce a family of classification models, scE2G, to predict enhancer–gene regulation.”

Then look for missing context

Context layer

What the story left out

Important study details the story did not include.

The paper benchmarks scE2G against CRISPR perturbations, fine-mapped eQTLs, and GWAS variant–gene associations and reports state-of-the-art performance across several cell types and perturbation categories.
The story includes a performance-oriented claim, but it frames performance around dataset size and sequencing depth rather than the benchmark sources and scope described in the abstract profile. The specific CRISPR/eQTL/GWAS benchmarking element is not accurately reflected.
From secondary_data benchmarking
For the INPP4B and IL15 lymphocyte-count examples, the abstract reports computational nomination but does not report experimental validation or quantitative support for those specific links.
The story caveats do not clearly state that these specific variant-to-gene links are nominations lacking experimental validation in the abstract-level evidence, which is important for interpreting the strength of the claim.
From secondary_data

3 things the story did carry across

The paper introduces scE2G as classification models predicting enhancer–gene regulatory interactions from single-cell ATAC-seq or multiomic RNA+ATAC features, trained on CRISPR perturbation ground truth with >10,000 evaluated element–gene pairs.
The paper applies scE2G to build enhancer–gene regulatory maps in heterogeneous tissues and to interpret noncoding complex-trait variants, nominating regulatory interactions involving INPP4B and IL15 for lymphocyte count.
The paper frames enhancer–gene mapping in specific cell types as important but challenging, motivating single-cell prediction methods.

Then read the study layer

Study layer

Study at a glance

Scan the study first. Expand only the parts you want to inspect.

Pieces of work

Evidence read

study summary

Lead result

in silico

1Lead resultin silicoIntroduce scE2G, a family of classification models that predict enhancer–gene regulatory interactions from single-cell ATAC-seq or multiome (RNA+ATAC) features, trained on large CRISPR perturbation ground truth.Supervised classification (scE2G) trained on CRISPR perturbation labelsExpand

In plain English

The paper introduces scE2G, a family of supervised classification models that predict enhancer–gene regulatory interactions using features derived from single-cell ATAC-seq or multiomic (RNA+ATAC) data; the models are trained on a CRISPR perturbation dataset comprising over 10,000 evaluated element–gene pairs and are benchmarked against genetic and perturbation-derived truth sets and applied to map regulatory interactions and interpret noncoding trait-associated variants.

Key findings

The authors introduce scE2G, a family of classification models to predict enhancer–gene regulatory interactions from single‑cell data.
scE2G models use features from single‑cell ATAC‑seq or multiomic RNA+ATAC data and are trained on a CRISPR perturbation dataset comprising over 10,000 evaluated element–gene pairs.>10,000 element–gene pairs used as training labels (reported in abstract)

“Here we introduce a family of classification models, scE2G, to predict enhancer–gene regulation.”

2secondary dataBenchmark scE2G against multiple external/orthogonal validation sources (CRISPR perturbations, fine-mapped eQTLs, GWAS variant–gene links) and show state-of-the-art predictive performance across cell types and perturbation categories.secondary data benchmarkingExpand

In plain English

The authors benchmark scE2G classification models for predicting enhancer–gene regulatory interactions using secondary datasets: CRISPR perturbation outcomes, fine-mapped eQTLs, and GWAS variant–gene associations. Models are trained on a CRISPR perturbation dataset with >10,000 evaluated element–gene pairs and are reported to achieve state-of-the-art predictive performance across multiple cell types and categories of perturbations.

Key findings

When benchmarked against CRISPR perturbations, fine-mapped eQTLs, and GWAS variant–gene associations, scE2G models are reported to achieve state-of-the-art performance for predicting enhancer–gene regulatory interactions across multiple cell types and perturbation categories.

“We benchmark scE2G models against CRISPR perturbations, fine-mapped expression quantitative trait loci and genome-wide association study variant–gene associations and demonstrate state-of-the-art performance at prediction tasks across several cell types and categories of perturbations.”

What this piece can’t prove

Abstract does not describe evaluation procedures (e.g., held-out test selection, cross-validation, or measures taken to prevent data leakage) needed to fully assess benchmarking rigor.

1 further detail could not be confirmed from the summary.

3secondary dataApply scE2G to build enhancer–gene maps in heterogeneous tissues and use them to interpret noncoding complex-trait variants, nominating specific regulatory links (for example INPP4B and IL15 to lymphocyte count).secondary dataExpand

In plain English

The authors applied the scE2G classification models (trained on a CRISPR perturbation dataset of >10,000 element–gene pairs) to single-cell ATAC-seq or multiomic RNA+ATAC datasets to construct enhancer–gene regulatory interaction maps in heterogeneous tissues and used these maps to interpret noncoding complex-trait variants, nominating regulatory links connecting INPP4B and IL15 to lymphocyte count.

Key findings

Application of scE2G to single-cell chromatin/multiome data produced enhancer–gene regulatory interaction maps in heterogeneous tissues and was used to interpret noncoding complex-trait variants.
Using these predicted maps, the authors nominate regulatory interactions linking INPP4B and IL15 to lymphocyte count.

“We apply scE2G to build maps of enhancer–gene regulatory interactions in heterogeneous tissues and interpret noncoding variants associated with complex traits, nominating regulatory interactions linking INPP4B and IL15 to lymphocyte count.”

What this piece can’t prove

Abstract provides limited methodological and dataset detail for the downstream mapping and variant-interpretation analyses (e.g., which tissue datasets were used, sample sizes, or thresholds for nomination).
The abstract does not report experimental validation or quantitative support for the nominated INPP4B and IL15 links to lymphocyte count.

Finally, the search trail

Method layer