Ahead of International Women’s Day, check out this collection on the rewards and challenges that face women in science, including news and research covering fresh approaches to women’s health, the motherhood penalty, and more.
go.nature.com/3OKCfhE
We still have some work to do before we go to peer review, so are happy to get any user feedback!
Things we are working on:
- test it on more multi-omics datasets
- figure out a way to auto-tune the weighting of different omics layers
- provide some sort of widget to faciliate exploratory analysis
Check out our new preprint describing the extensions to SpatialLeiden clustering:
- it's atlas scale (supporting intergration, and fast)
- it's 3D ready (it accepts connections over serial sections)
- it's multiomic ready (it accepts multiple omics layers)
www.biorxiv.org/content/10.6...
Some things we are looking into:
- tuning the resolution for each multi omics layer is work. SpatialLeiden is fast, so parameter scans are not bad, but we are looking for more elegant solutions.
- we are still looking for a very good spatial multiomics dataset for demonstrating advantage
Have you heard of @elixir-europe.org? It's a life sciences research infrastructure, bringing together #bioinformatics resources across Europe to store, analyse, share and reuse data.
They are behind UniProt, ChEMBL, PDBe, ENA, bio.tools,...
Read more here: f1000research.com/articles/14-....
Due to popular demand, we just extended documentation of SpatialLeiden to include guides for multi-sample, 3D and multi-omics spatial clustering: spatialleiden.readthedocs.io/latest/guide....
It's still remains as fast and intuitive as standard Leiden clustering.
A preprint is on the way...
You can find the documentation and user guides here: ovrlpy.readthedocs.io/latest/
All code is up on Github: github.com/HiDiHlabs/ov...
Happy to hear about your thoughts and experiences using the tool!
Our manuscript describing how cell overlaps in tissues sections affects spatial transcriptomics is now live on Nature Biotechnology: www.nature.com/articles/s41....
With it be provide the ovrlpy python tool to identify signals of overlapping cells in tissues, and assigning overlap scores to cells.
Sainsc is a new tool for efficient whole organism spatial transcriptomics data analysis at the nanometre scale. It works with Stereo-seq, VisiumHD, Xenium and Nova-ST. Looking forward to see how it works with upcoming platforms (e.g. Illumina).
github.com/HiDiHlabs/sa...
Spatial transcriptomics - helping authors wow reviewers with pretty pictures (... and interesting spatial biology).. and making it possible for bioinformaticians to submit journal cover art! Our Sainsc tool is featured as the back cover for Small Methods: onlinelibrary.wiley.com/doi/10.1002/...
And our tireless worker bees who dedicated every Wednesday evening for 18 months for this project: Jieran Sun, Kirti Biharie, Peiying Cai, Niklas Müller-Bötticher, Paul Kiessling, Meghan A. Turner, Søren H. Dam, Florian Heyl. 🙏
I forgot the most important bit! This would not have been possible without the wonderful and amazing SpaceHack community. Special shout outs to Brain Long, Ahmed Mahfouz and @markrobinsonca.bsky.social for their work, ideas and resources...
We wrap up our analysis framework into SACCELERATOR, github.com/SpatialHacka..., and present two case studies on brain and colon cancer datasets to remind people that having domain expertise and performing interactive analysis is more important than picking the "best" tool.
General findings:
- benchmarking studies can be inconsistent
- methods do not scale to new datasets
- "ground truths" do not consider granularity of biological features
- looking at the differences between methods (entropy) is informative
- method performance doesn't match subjective importance
Did you ever notice discrepancies in benchmarking of bioinformatics tools? We did too! Setting out to benchmark spatial clustering methods, we encountered major reproducibility issues in previous benchmarks and questionable "ground truths".
More in our preprint: www.biorxiv.org/content/10.1...
Accurate cell segmentation remains a major issue for spatial transcriptomics.
Elyas Heidari and colleagues from the Gerstung, Pe'er and Stegle Labs released Segger, a new algorithm that uses GNN to model both transcripts and cells. More details in their preprint: www.biorxiv.org/content/10.1...
🧬If you work with @10xgenomics.bsky.social Xenium, this might be of your interest.
Our new study on Xenium exploration best practice and comparing it with other commercial platforms is now out in @naturemethods.bsky.social
www.nature.com/articles/s41...
We also released the ovrl.py tool to identify areas of cell overlaps in imaging-based spatial transcriptomic data (e.g. #Xenium, #MERSCOPE, #cosMX).
GitHub: github.com/HiDiHlabs/ov...
Have you been bothered that practically every #SpatialTranscriptomic analysis tool assumes data to be 2D? Well, you were right - overlapping cells are a really big problem. Check out our report on bioRxiv: 2D, or not 2D? Investigating Vertical Signal Integrity of Tissue Slices. t.ly/CQX8K
We also sneak in a python implementation for the spatially aware MULTISPATI-PCA. We think this should pretty much replace "non-spatial" PCA for all spatial modalities in most cases.
GitHub: github.com/HiDiHlabs/mu...
RtD: multispaeti.readthedocs.io/1.0.0/genera...
SpatialLeiden integrates with the #scverse by leveraging #scanpy and #anndata but can also be used independently.
GitHub: github.com/HiDiHlabs/sp...
RtD: spatialleiden.readthedocs.io/stable/
SpatialLeiden is out in #GenomeBiology: t.ly/WQppF. It models both gene expression and spatial information for clustering. This allows you to perform e.g. spatially aware cell typing or domain identification. It's fast, scalable and intuitive for single-cell researchers. Try it out!
The HISTOMAP projects aims to use AI and spatial transcriptomics data to accelerate biomarker detection for bladder cancer. We will work closely with the clinic, pathology, and regulatory experts to investigate how best to get these models into clinical routine.
Do you also think that #SpatialTranstriptomics and #AI will give rise to #DigitalPathology 2.0? We have an open PhD position for the HISTOMAP (Histology-based Molecular Analysis Platform) project to accelerate biomarker detection. Please apply here: www.bihealth.org/en/notices/b....
Spread the word!
I tried Sainsc for our MERSCOPE data and Naveed, it works wonderfully! Importantly, the documentation is excellent. Thank you so much for the clear tutorial and explanations. This is truly a life saver for my analysis at the moment. Props to the team, this is awesome work!
Everyone knows by now that 42 is the answer to the ultimate question of life, the universe, and everything. But as I turn 42 today, I’ve realised something profound: it’s not the answer that matters - it’s the prompt!
Am I really the first to repost Mats Nilsson? What an honour!
Hello world. Our lab has a new home here on Bluesky! We will post about spatial omics, science, nice papers and lab activities. The account will be collectively managed by lab members. Happy spatial-omics everyone!
The calm before the storm... #SpaceHack starts tomorrow! Looking forward to 3 intense days of addressing unsolved data analysis problems for #SpatialTranscriptomics.
More here: spatialhackathon.github.io
SpatialLeiden adds spatial support to the Leiden clustering algrithm. It has 3 parameters. Two we know from scRNAseq analysis (resolution and num of neighbors), and the other is the weight of the spatial info. We thin it is the natural choice for spatial clustering for the single-cell community!
