Kalin Nonchev
@nonchev.bsky.social
PhD at ETH Zurich, machine learning and biomedical data https://kalinnonchev.github.io
Older spot-level spatial transcriptomics datasets shouldn't be forgotten now that new single-cell methods exist. 🧬
Instead of discarding this rich resource, we can bridge the gap.
DeepSpot2Cell helps bridge the gap 👇
Instead of discarding this rich resource, we can bridge the gap.
DeepSpot2Cell helps bridge the gap 👇
October 1, 2025 at 3:28 PM
Older spot-level spatial transcriptomics datasets shouldn't be forgotten now that new single-cell methods exist. 🧬
Instead of discarding this rich resource, we can bridge the gap.
DeepSpot2Cell helps bridge the gap 👇
Instead of discarding this rich resource, we can bridge the gap.
DeepSpot2Cell helps bridge the gap 👇
Internship Opportunity: Multimodal AI Research Scientist at the Biomedical Informatics Group at ETH Zurich 🚀
Interested in working at the intersection of computational pathology, spatial transcriptomics, LLM representation learning, and tissue generation?
Interested in working at the intersection of computational pathology, spatial transcriptomics, LLM representation learning, and tissue generation?
August 18, 2025 at 8:41 PM
Internship Opportunity: Multimodal AI Research Scientist at the Biomedical Informatics Group at ETH Zurich 🚀
Interested in working at the intersection of computational pathology, spatial transcriptomics, LLM representation learning, and tissue generation?
Interested in working at the intersection of computational pathology, spatial transcriptomics, LLM representation learning, and tissue generation?
🚀 Excited to share that we've generated the largest digital spatial transcriptomics dataset using DeepSpot - over 56 million spatial transcriptomics spots from 3 780 TCGA samples across skin melanoma, renal cell carcinoma, lung adenocarcinoma, and lung squamous cell carcinoma cohorts. #pathology
May 12, 2025 at 5:07 AM
🚀 Excited to share that we've generated the largest digital spatial transcriptomics dataset using DeepSpot - over 56 million spatial transcriptomics spots from 3 780 TCGA samples across skin melanoma, renal cell carcinoma, lung adenocarcinoma, and lung squamous cell carcinoma cohorts. #pathology
By extending our recent deep learning method, DeepSpot, to support 10x Genomics Xenium data, we significantly improved single-cell gene expression predictions in patients with Inflammatory Bowel Disease. It is exciting to see its performance validated in an independent evaluation!
March 20, 2025 at 6:36 AM
By extending our recent deep learning method, DeepSpot, to support 10x Genomics Xenium data, we significantly improved single-cell gene expression predictions in patients with Inflammatory Bowel Disease. It is exciting to see its performance validated in an independent evaluation!
First place award at the Autoimmune Disease Machine Learning Challenge organized by the @broadinstitute.org and CrunchDAO. Our approach outperformed competitors worldwide in predicting single-cell spatial transcriptomics from H&E images. 🎉
March 20, 2025 at 6:36 AM
First place award at the Autoimmune Disease Machine Learning Challenge organized by the @broadinstitute.org and CrunchDAO. Our approach outperformed competitors worldwide in predicting single-cell spatial transcriptomics from H&E images. 🎉
9/12 The TCGA spatial transcriptomics dataset, containing over 37 million spots, provides unique insights into the molecular landscapes of cancer tissues. It also sets a benchmark for evaluating and developing new spatial transcriptomics models. 🌍
February 24, 2025 at 7:46 PM
9/12 The TCGA spatial transcriptomics dataset, containing over 37 million spots, provides unique insights into the molecular landscapes of cancer tissues. It also sets a benchmark for evaluating and developing new spatial transcriptomics models. 🌍
8/12 DeepSpot outperformed previous models or matched bulk-RNA seq performance in tumor type classification. 🧬
February 24, 2025 at 7:46 PM
8/12 DeepSpot outperformed previous models or matched bulk-RNA seq performance in tumor type classification. 🧬
7/12 Using DeepSpot, we generated the largest spatial transcriptomics dataset - over 37 million spots from 1792 TCGA slides across skin melanoma and kidney cancer. To assess out-of-distribution performance we aggregated the gene expression per slide and compared it to bulk-RNA.
February 24, 2025 at 7:46 PM
7/12 Using DeepSpot, we generated the largest spatial transcriptomics dataset - over 37 million spots from 1792 TCGA slides across skin melanoma and kidney cancer. To assess out-of-distribution performance we aggregated the gene expression per slide and compared it to bulk-RNA.
6/12 DeepSpot overcomes spatial transcriptomics gene sensitivity. Higher resolution often reduces gene sensitivity and increases spot contamination, leading to misleading results. DeepSpot successfully recovered low-quality samples that would’ve been discarded.
February 24, 2025 at 7:46 PM
6/12 DeepSpot overcomes spatial transcriptomics gene sensitivity. Higher resolution often reduces gene sensitivity and increases spot contamination, leading to misleading results. DeepSpot successfully recovered low-quality samples that would’ve been discarded.
5/12 DeepSpot enables digital spatial transcriptomics analysis, allowing de novo clustering, marker gene identification, and pathway discovery based on predicted gene expression. 🔍
February 24, 2025 at 7:45 PM
5/12 DeepSpot enables digital spatial transcriptomics analysis, allowing de novo clustering, marker gene identification, and pathway discovery based on predicted gene expression. 🔍
4/12 DeepSpot substantially improved gene correlations across multiple datasets from patients with metastatic melanoma, kidney, lung, and colon cancers, outperforming previous state-of-the-art models. 🚀
February 24, 2025 at 7:45 PM
4/12 DeepSpot substantially improved gene correlations across multiple datasets from patients with metastatic melanoma, kidney, lung, and colon cancers, outperforming previous state-of-the-art models. 🚀
3/12 DeepSpot uses a deep-set neural network to model transcriptomic spots as bags of sub-spots (Visium). It enhances resolution by learning sub-spot contributions and integrates tissue context by pooling neighboring spots to learn the tissue landscape.
February 24, 2025 at 7:45 PM
3/12 DeepSpot uses a deep-set neural network to model transcriptomic spots as bags of sub-spots (Visium). It enhances resolution by learning sub-spot contributions and integrates tissue context by pooling neighboring spots to learn the tissue landscape.
2/12 For each spot, we create a bag of sub-spots by dividing it into sub-tiles that capture the local morphology and a bag of neighboring spots to represent the global tissue environment. A pretrained pathology foundation model extracts tile features, which are input to DeepSpot.
February 24, 2025 at 7:44 PM
2/12 For each spot, we create a bag of sub-spots by dividing it into sub-tiles that capture the local morphology and a bag of neighboring spots to represent the global tissue environment. A pretrained pathology foundation model extracts tile features, which are input to DeepSpot.
1/12 DeepSpot leverages advanced pathology foundation models and multi-level tissue context to accurately predict gene expression directly from H&E images. It is trained on paired spatial transcriptomics data sequenced with @10xgenomics.bsky.social
Visium. 💻✨
Visium. 💻✨
February 24, 2025 at 7:44 PM
1/12 DeepSpot leverages advanced pathology foundation models and multi-level tissue context to accurately predict gene expression directly from H&E images. It is trained on paired spatial transcriptomics data sequenced with @10xgenomics.bsky.social
Visium. 💻✨
Visium. 💻✨
How can we predict spatial transcriptomics from histology images to enable simple, affordable, and reliable analysis of spatially resolved gene expression in routine clinical use? 🤔 Introducing DeepSpot – a deep learning model designed to tackle this! 🧵👇 www.medrxiv.org/content/10.1...
February 24, 2025 at 7:43 PM
How can we predict spatial transcriptomics from histology images to enable simple, affordable, and reliable analysis of spatially resolved gene expression in routine clinical use? 🤔 Introducing DeepSpot – a deep learning model designed to tackle this! 🧵👇 www.medrxiv.org/content/10.1...
9/12 The TCGA spatial transcriptomics dataset, containing over 37 million spots, provides unique insights into the molecular landscapes of cancer tissues. It also sets a benchmark for evaluating and developing new spatial transcriptomics models. 🌍
February 24, 2025 at 7:36 PM
9/12 The TCGA spatial transcriptomics dataset, containing over 37 million spots, provides unique insights into the molecular landscapes of cancer tissues. It also sets a benchmark for evaluating and developing new spatial transcriptomics models. 🌍
8/12 DeepSpot outperformed previous models or matched bulk-RNA seq performance in tumor type classification. 🧬
February 24, 2025 at 7:35 PM
8/12 DeepSpot outperformed previous models or matched bulk-RNA seq performance in tumor type classification. 🧬
7/12 Using DeepSpot, we generated the largest spatial transcriptomics dataset - over 37 million spots from 1792 TCGA slides across skin melanoma and kidney cancer. To assess out-of-distribution performance we aggregated the gene expression per slide and compared it to bulk-RNA.
February 24, 2025 at 7:35 PM
7/12 Using DeepSpot, we generated the largest spatial transcriptomics dataset - over 37 million spots from 1792 TCGA slides across skin melanoma and kidney cancer. To assess out-of-distribution performance we aggregated the gene expression per slide and compared it to bulk-RNA.
6/12 DeepSpot overcomes spatial transcriptomics gene sensitivity. Higher resolution often reduces gene sensitivity and increases spot contamination, leading to misleading results. DeepSpot successfully recovered low-quality samples that would’ve been discarded.
February 24, 2025 at 7:34 PM
6/12 DeepSpot overcomes spatial transcriptomics gene sensitivity. Higher resolution often reduces gene sensitivity and increases spot contamination, leading to misleading results. DeepSpot successfully recovered low-quality samples that would’ve been discarded.
5/12 DeepSpot enables digital spatial transcriptomics analysis, allowing de novo clustering, marker gene identification, and pathway discovery based on predicted gene expression. 🔍
February 24, 2025 at 7:34 PM
5/12 DeepSpot enables digital spatial transcriptomics analysis, allowing de novo clustering, marker gene identification, and pathway discovery based on predicted gene expression. 🔍
4/12 DeepSpot substantially improved gene correlations across multiple datasets from patients with metastatic melanoma, kidney, lung, and colon cancers, outperforming previous state-of-the-art models. 🚀
February 24, 2025 at 7:33 PM
4/12 DeepSpot substantially improved gene correlations across multiple datasets from patients with metastatic melanoma, kidney, lung, and colon cancers, outperforming previous state-of-the-art models. 🚀
3/12 DeepSpot uses a deep-set neural network to model transcriptomic spots as bags of sub-spots (Visium). It enhances resolution by learning sub-spot contributions and integrates tissue context by pooling neighboring spots to learn the tissue landscape.
February 24, 2025 at 7:33 PM
3/12 DeepSpot uses a deep-set neural network to model transcriptomic spots as bags of sub-spots (Visium). It enhances resolution by learning sub-spot contributions and integrates tissue context by pooling neighboring spots to learn the tissue landscape.
2/12 For each spot, we create a bag of sub-spots by dividing it into sub-tiles that capture the local morphology and a bag of neighboring spots to represent the global tissue environment. A pretrained pathology foundation model extracts tile features, which are input to DeepSpot.
February 24, 2025 at 7:33 PM
2/12 For each spot, we create a bag of sub-spots by dividing it into sub-tiles that capture the local morphology and a bag of neighboring spots to represent the global tissue environment. A pretrained pathology foundation model extracts tile features, which are input to DeepSpot.
1/12 DeepSpot leverages advanced pathology foundation models and multi-level tissue context to accurately predict gene expression directly from H&E images. It is trained on paired spatial transcriptomics data sequenced with @10xGenomics
Visium. 💻✨
Visium. 💻✨
February 24, 2025 at 7:32 PM
1/12 DeepSpot leverages advanced pathology foundation models and multi-level tissue context to accurately predict gene expression directly from H&E images. It is trained on paired spatial transcriptomics data sequenced with @10xGenomics
Visium. 💻✨
Visium. 💻✨