Lightnews — Scholar-powered news

Marco Varrone

@marcovarrone.bsky.social

750 followers 110 following 35 posts

Postdoc in Computational Biology at @CirielloLab @dbc_unil. Machine Learning for spatial omics. https://batcheffect.com

batcheffect.com

Posts Media Videos Starter Packs

Pinned

Marco Varrone @marcovarrone.bsky.social · Dec 8

Excited that our paper "CellCharter reveals spatial cell niches associated with tissue remodeling and cell plasticity" is finally out!

CellCharter is a computational framework to identify and study cellular niches in large-scale spatial transcriptomics, proteomics, and even epigenomics data.

CellCharter reveals spatial cell niches associated with tissue remodeling and cell plasticity - Natu...

CellCharter is a flexible, platform-agnostic method for identifying cell niches in spatially resolved data. Analysis of lung cancers demonstrates the importance of considering spatial information, exe...

www.nature.com

1 1 4

Marco Varrone @marcovarrone.bsky.social · Aug 15

Great work @grst.bsky.social ! I remember there were some differences between DESeq2 (R) and pyDESeq2 (Python).

Do you think it's safe to assume that pyDESeq2 is also the best approach for the simple case? :)

Marco Varrone @marcovarrone.bsky.social · Jul 15

And for anyone who has considered contributing to an open source package: don't be scared to propose changes.

Even if it's not a complete and perfect solution, whoever is maintaining the package will help you in get to the right solution and they will be incredibly grateful.

Marco Varrone @marcovarrone.bsky.social · Jul 15

For people like me who don't have a team behind a package like CellCharter, contributions like these mean a lot. So thank you Lukas :)

And congratulations, it's not always easy to jump into an existing codebase and propose changes.

1 1 1

Marco Varrone @marcovarrone.bsky.social · Jul 15

RCS measures how large a cell niche (aka spatial domain aka spatial cluster) is compared to what would be normally expected.

This pushed me to completely rewrite the system for generating and plotting boundaries for cell niches.
The new system is now more efficient, consistent, and visually clear.

1 1 1

Marco Varrone @marcovarrone.bsky.social · Jul 15

A few weeks ago, we released CellCharter v0.3.5.

Among some bug fixes, we received our first contribution from an external contributor: @loggas.bsky.social !

He designed a new metric called Relative Component Size (RCS).

GitHub - CSOgroup/cellcharter: A Python package for the identification, characterization and comparison of spatial clusters from spatial -omics data.

A Python package for the identification, characterization and comparison of spatial clusters from spatial -omics data. - CSOgroup/cellcharter

github.com

2 1 5

Reposted by Marco Varrone

Giovanni Ciriello @ciriellolab.bsky.social · Mar 18

After 6 (healthy) months of hiatus from social media, first post with a brand new conference announcement!! 🔥COMPUTATIONAL BIOLOGY IS EVERYWHERE! Join us in Lausanne for the first international Computational Biology Symposium! 18-19 September 2025 Registrations are OPEN! cbiosymposium.unil.ch

Computational Biology Symposium

cbiosymposium.unil.ch

2 11 19

Marco Varrone @marcovarrone.bsky.social · Mar 12

By the way, here in Switzerland the public defense is meant for the general public, so the presentation is understandable by anyone, and I even sneaked some jokes inside.

1 1

Marco Varrone @marcovarrone.bsky.social · Mar 12

Last year I completed my PhD and my father recorded my public defense 📹

So I decided to share it!

I talked about the communities of cells in our bodies and how we develop algorithms to study them.

Hope you enjoy it!

Computational models to characterize spatial intra-tumor heterogeneity | Marco Varrone

YouTube video by Marco Varrone

youtu.be

1 2

Reposted by Marco Varrone

Jacob Schreiber @jmschreiber91.bsky.social · Jan 6

The only sort of gatekeeper I wish we saw more of were those who wanted better software. Who cares if your model performs the best if it's unusable, or if your pipeline aligs reads faster if it only works on your local work computer? Or if there's no documentation on how to use any of it.

1 3 29

Marco Varrone @marcovarrone.bsky.social · Dec 19

The scverse ecosystem is one of the best things that could have ever happened to the single-cell and spatial fields.

Smoother computational analyses mean faster and better research

scverse @scverse.bsky.social · Dec 19

Closing 2024, scverse celebrates growth and collaboration! Huge thanks to everyone on this journey and our core team & board (@fabiantheis.bsky.social, @francescafinotello.bsky.social, @steglelab.bsky.social, Nir Yosef, @panosroussos.bsky.social, @robp.bsky.social). Here's to an inspiring 2025!

1 2 14

Marco Varrone @marcovarrone.bsky.social · Dec 18

Contributing to this project has been very cool!

Quick tip for anyone interested in cell segmentation: I started with StarDist but it failed to detect nuclei in samples with regions of very different cell density.

Switched to DeepCell Mesmer and it worked like a charm!

Elisa Oricchio @elisaoricchio.bsky.social · Dec 16

Just before the holidays, here is the latest from the Oricchio lab. In collaboration with the hemato-oncology team, we established a new model to test therapies on lymphoma patient samples by preserving the tissue composition and architecture of the original tumors. www.nature.com/articles/s41...

Development of patient-derived lymphomoids with preserved tumor architecture for lymphoma therapy screening - Nature Communications

Choosing the most effective anti-cancer therapy is difficult due to the high degree of tumour cell and tumour microenvironment heterogeneity. Here, the authors develop a platform for the generation of...

www.nature.com

1 8

Reposted by Marco Varrone

khrovatin.bsky.social @khrovatin.bsky.social · Dec 16

To bring to light data science topics that usually don’t make it into publications I started a blog on this topic: hrovatin.github.io By interviewing different researchers, I plan to find out what is going on in the community.

Karin Hrovatin

Data science blog on topics that don’t get published.

hrovatin.github.io

1 6 10

Marco Varrone @marcovarrone.bsky.social · Dec 12

3. How spatial omics can identify cell communities and true cell interactions to understand what is really happening.

4. How discoveries driven by spatial omics will lead to new effective therapies.

1 2

Marco Varrone @marcovarrone.bsky.social · Dec 12

1. How two types of cells, fibroblasts and macrophages, can cooperate to support cancer.

2. Why some of the therapies targeting only one of those cell types failed in clinical trials.

1 2

Marco Varrone @marcovarrone.bsky.social · Dec 12

We are used to thinking that, when an organism has cancer, all other cells will fight it.

We are now realizing that this is not true, and some of these cells may even help it.

In the fourth Batch Effect post I talked about 4 aspects:

The hidden allies of cancer

How spatial omics is bringing normal cell interactions into focus

www.batcheffect.com

1 3

Marco Varrone @marcovarrone.bsky.social · Dec 5

🔬 Spatial omics: revolutionary breakthrough or just incremental progress?

In the third Batch Effect blog post, I compared this emerging technology to the historic hype around interferon therapy.

Is this time going to be different?

www.batcheffect.com/p/breakthrou...

Breakthrough or incremental progress?

Why spatial omics is revolutionary for biology and medicine

www.batcheffect.com

2 5

Marco Varrone @marcovarrone.bsky.social · Nov 27

What do fruit pies 🥧, lego sets🧩, and cities🏙️ have in common?

They are all metaphors used to describe spatial transcriptomics.

In the second Batch Effect blog post I explored the best metaphor to make anyone understand what is spatial transcriptomics

www.batcheffect.com/p/the-best-m...

The best metaphor for spatial transcriptomics

How everyone can understand what is spatial transcriptomics

www.batcheffect.com

2 11

Marco Varrone @marcovarrone.bsky.social · Nov 25

Bonus tip: if you are considering creating a package to be included within the scverse ecosystem, the cookiecutter template already sets up 5 of the 8 mandatory requirements.

Marco Varrone @marcovarrone.bsky.social · Nov 25

A common problem with templates and frameworks is that they tend to restrict what you can do to keep you consistent.

But this is not the case here.

I used the scverse template even for projects that are not strictly related to single-cell or computational biology.

Marco Varrone @marcovarrone.bsky.social · Nov 25

It sets up beautiful documentation generated directly from your code like in the picture.

Finally, it gives automatic actions for publishing and updating the package.

Marco Varrone @marcovarrone.bsky.social · Nov 25

Every time you push code to your main GitHub branch, scverse’s cookiecutter automatically runs tests.

Even better, it supports testing with pre-release dependencies!

This means you can detect breaking changes in upcoming dependency updates before they're officially released.

Marco Varrone @marcovarrone.bsky.social · Nov 25

If you are considering creating a Python package I strongly suggest cookiecutter-scverse by @scverse.bsky.social

It’s a template that already incorporates many useful features.

github.com/scverse/cook...

1 2 11

Marco Varrone @marcovarrone.bsky.social · Nov 24

Same as what was said about React, Vue may be a bit overkill unless you want 100% customization.

But if you are interested or you want to use it as an excuse to learn Vue.js, there is also an official project called VuePress that is focused on showing content using Markdown

Marco Varrone @marcovarrone.bsky.social · Nov 21

You mean installing VS Code? It should be much easier than something like the Linux subsystem.

You just need to download the VS Code file and click install

Marco Varrone @marcovarrone.bsky.social · Nov 21

I do! I find it not that different from JupyterLab.

It's just nice to have everything in one place since I tend to mix scripts with notebooks.

1 1