Every voiceis a nodein the network.
NetREm (Network Regression Embeddings) finds which predictors work together to shape an outcome. It was inspired by a biological problem and built to reveal transcription factor coordination in cells of organisms, and it works in any field where predictors interact in a network. This page maps how the people using it connect 🤓🌐. Tell us what NetREm did for your thesis, your lab, your work, or your class, and your story joins the graph 📊.
Hover a coloured node to read a voice
Community pulse
Updated live as students and researchers add their experiences 🌎.
How people rate NetREm
Who's here
What they use it for
Testimonials
In their own words: how NetREm fits into real research and coursework.
Heard about NetREm in a course or seminar?
You don't need to have used the tool yet. Please tell us what you learned, what clicked, and what didn't, so the next talk is even better 😄.
Insights dashboard
What the community is saying, in numbers and words. Everything here updates as new testimonials and seminar feedback arrive.
Word cloud
Bigger words appear more often. Please click a word to find testimonials that mention it.
Responses over time
New testimonials and seminar responses per month
Testimonial rating by role
Average stars, with the number of reviews
Seminar scores by session type
Average overall, clarity, and relevance, out of 5
Will attendees try NetREm?
From all seminar and course responses
Fields and domains
Where people apply NetREm, from testimonials
Where sessions happened
Locations attendees reported
Case studies
Worked applications of NetREm, each as a PDF you can read online or download.
About NetREm
NetREm (Network Regression Embeddings) is an open-source computational framework for finding which transcription factors work together to regulate target genes in a specific cell type.
It combines single-cell gene expression with prior knowledge of protein–protein interactions among TFs, using network-constrained regularization to produce embeddings that quantify TF–TF coordination. It has been applied to immune cell sub-types, neuronal and glial cells, Schwann cells, and Alzheimer's disease.
Under the hood, NetREm is a general-purpose method: it suits any problem where predictors form subnetworks that jointly influence an outcome. See it in other fields.
Khullar S, Huang X, Ramesh R, Svaren J, Wang D. NetREm: Network Regression Embeddings reveal cell-type transcription factor coordination for gene regulation. bioRxiv 2023.10.25.563769. doi:10.1101/2023.10.25.563769
- Start with expressionSingle-cell gene expression for one cell type: candidate TFs and a target gene.
- Add the interaction priorA protein–protein interaction network among TFs guides the regression.
- Regularized regressionNetwork and sparsity hyperparameters select the TFs that matter for each gene.
- Read the embeddingsEmbeddings reveal which TFs coordinate, producing TF–TG and TF–TF networks.
Built for any network of predictors
Gene regulation is where NetREm started, not where it stops. Whenever many predictors are connected to each other and together drive one outcome, NetREm can find which ones matter and which ones work as a team.
Try the two dials
A simplified illustration of what NetREm's two hyperparameters do. It isn't the real model fit.
Higher β lets linked predictors share strength, so teams of connected predictors rise together.
Higher α keeps only the strongest predictors and sets the rest to zero.
Selected predictors
Coordination scores
Every pair of selected predictors gets a score from −100 to +100, so the final network is fully connected. Positive (green) means both predictors push the outcome the same way: two activators or two repressors. Negative (red) means they push opposite ways: one activator and one repressor. Outlined cells were links in your prior network; the rest are newly inferred.
The model at a glance
Same recipe in every field: an outcome, its predictors, and what's already known about how those predictors connect.
What NetREm balances
- Fit. How well a weighted combination of predictors explains the outcome y.
- Network penalty. Uses your prior network so that predictors known to interact are treated as related, not as strangers.
- Sparsity penalty. A lasso-style term that drops predictors with little effect, keeping results interpretable.
Schematic form. See the paper for NetREm's exact formulation and how the embeddings are built.
You bring
- An outcome measured across samples (a target gene's expression, a test score, a crop yield).
- Predictors measured on the same samples, often many more predictors than samples.
- A prior network listing which predictors are known or expected to interact, optionally weighted.
NetREm returns
- Selected predictors with signed coefficients: which ones drive the outcome, and whether each raises it (activator, +) or lowers it (repressor, −).
- Network regression embeddings that place predictors in a shared space.
- Coordination scores from −100 to +100 for every pair of selected predictors. Positive when both push the outcome the same way (two activators or two repressors), negative when they push opposite ways (an activator and a repressor). The result is a fully connected network: some links confirm your prior, others are new discoveries.
NetREm is a good fit when
Questions
Do I have to work in biology to use NetREm or leave a testimonial?
No. NetREm is general purpose: any field where predictors interact in a network and jointly affect an outcome can use it. If you applied it in neuroscience, ecology, finance, engineering, or anywhere else, we especially want to hear from you. Pick "Beyond biology" and add your field in the form.
I only heard a talk and haven't used NetREm. Can I still give feedback?
Yes, please do. The seminar feedback form is made for exactly that: tell us what you learned, what was unclear, and whether you might try it.
Is my email shown publicly?
Never. Your email goes only to the NetREm team. The public card shows your name, role, optional institution and field, rating, and testimonial.
Can I send feedback without posting it publicly?
Yes. In the last step, uncheck "Post my testimonial on the public wall". Your message and any suggestions then go privately to the team.
Can I edit what I posted?
On the live community board, opening the form again loads your submission and saving updates your existing card. If you need something removed, contact the NetREm team.
I'm a student new to this. Where do I start with NetREm?
Start with the GitHub repository's README and the paper's overview figure, then try the dials above to build intuition for β and α. If something in the docs confused you, tell us in the private suggestions box.
Can I collaborate on NetREm or contribute a case study?
Yes, please. See Meet the creator for ways to get in touch. Case studies from new fields are especially welcome.
How should I cite NetREm?
Use the citation in the About section, which has a copy button, or the published paper linked there.
Team inbox
Only you can see this section. Names, emails, private suggestions, and seminar feedback from everyone who submitted.
| Name | Role | Institution | Project link | Rating | On wall | Updates | Private suggestions | Submitted |
|---|
Seminar and course feedback
| Name | Role | Session | Overall | Clarity | Relevance | Will try | Most valuable | Unclear | Apply to | Wants more on | Questions | Featured | Updates | Submitted |
|---|