Christoph Bock Lab @ CeMM & MedUni Vienna

Nice example of "garbage-in, garbage out" behavior of CellWhisperer. We could probably add checks to mark meaningless queries. But for now we rely on cooperative users, and CellWhisperer has a clear warning: "Please keep in mind that CellWhisperer is an AI system and may produce misleading results."

This is figure 1, which gives an overview of the CellWhisperer multimodal AI for natural-language analysis of transcriptome data.

A paper in Nature Biotechnology presents CellWhisperer, which uses multimodal learning of transcriptomes and text to answer questions about single-cell RNA-sequencing data. go.nature.com/3XqzItR 🧬 🧪

Multimodal learning enables chat-based exploration of single-cell data - Nature Biotechnology CellWhisperer uses multimodal learning of transcriptomes and text to answer questions about single-cell RNA-sequencing data.

Multimodal learning enables chat-based exploration of single-cell data - @bocklab.bsky.social @cemm.oeaw.ac.at @meduniwien.ac.at go.nature.com/3WPGJnW

🤝 Huge thanks to the team! Moritz Schaefer & Peter Peneder with Daniel Malzl, Salvo Lombardo, Mihaela Peycheva, Jake Burton, Anna Hakobyan, Varun Sharma, Thomas Krausgruber, Celine Sin, Jörg Menche, Eleni Tomazou, Christoph Bock. @cemm.oeaw.ac.at, @meduniwien.ac.at, @stanna-ccri.bsky.social (11/11)

🧬 CellWhisperer introduces a chat-based way to explore scRNA-seq data. By enabling natural language analysis, it bridges biologists and bioinformaticians—paving the way for AI-driven bioinformatics assistants. (10/11)

Multimodal learning enables chat-based exploration of single-cell data - Nature Biotechnology CellWhisperer uses multimodal learning of transcriptomes and text to answer questions about single-cell RNA-sequencing data.

Ready to talk to cells?
📖 Read the paper: doi.org/10.1038/s415...
🧬 Try the web app with public datasets: cellwhisperer.bocklab.org
🖥️ Analyze your own datasets: github.com/epigen/cellw...
(9/11)

📚 We trained on >1 million bulk & pseudo-bulk transcriptomes with textual annotations that we AI-curated from GEO & @CELLxGENE Census. Our training data is open source and useful for developing multimodal biomedical AI models and future bioinformatics research assistants. (8/11)

🪄 How does CellWhisperer work behind the scenes? We trained a multimodal AI that links transcriptomes and text, enabling free-text search and annotation of RNA profiles. And we connected this model to an LLM that we fine-tuned into a chat assistant for transcriptome data (7/11)

🚀 We also validated CellWhisperer’s chat-based analysis with conventional bioinformatics. CellWhisperer was >4x faster (and 10x cooler 😊). Our recommendation: Use CellWhisperer for dataset exploration – but statistics is still important to ensure rigor & reproducibility (6/11)

🆕 The CellWhisperer paper (doi.org/10.1038/s415...) includes several new analyses beyond our 2024 bioRxiv preprint (biorxiv.org/content/10.1...). For example, we used CellWhisperer for an AI-guided analysis of human organ development (5/11)

🔬 You can easily query large transcriptome datasets for your favorite biological process using CellWhisperer. Just open Tabula Sapiens (cellwhisperer.cemm.at/tabulasapiens/) or GEO (cellwhisperer.cemm.at/geo/) in CellWhisperer & type your query into the chat box – for example “infection” (4/11)

🔍 We investigate one of the identified cell clusters by selecting the cells & prompting CellWhisperer with ‘Describe these cells in detail’. This interactive workflow is enabled by seamless integration of the CellWhisperer AI chat box into a version of CELLxGENE Explorer (3/11)

⚙️ To get started, let’s find cells by typing into the CellWhisperer chat box. For example ‘Show me structural cells with immune functions’. CellWhisperer scores each transcriptome by how well it matches this textual query and colors by query match (red: high, blue: low) (2/11)

🗨️ Just published in Nature Biotechnology: Our CellWhisperer AI enables chat-based analysis of single-cell sequencing data. You can talk to your cells & figure out the biology without writing any computer code. Paper here: www.nature.com/articles/s41.... Annotated walkthrough in a thread below (1/11)

Multimodal learning enables chat-based exploration of single-cell data - Nature Biotechnology CellWhisperer uses multimodal learning of transcriptomes and text to answer questions about single-cell RNA-sequencing data.

Ready to talk to cells?
📖 Read the paper: www.nature.com/articles/s41...
🧬 Try the web app with public datasets: cellwhisperer.bocklab.org
🖥️ Analyze your own datasets: github.com/epigen/cellw...
(9/11)

📚 We trained on >1 million bulk & pseudo-bulk transcriptomes with textual annotations that we AI-curated from GEO & CELLxGENE Census. Our training data is open source and useful for developing multimodal biomedical AI models and future bioinformatics research assistants. (8/11)

🪄 How does CellWhisperer work behind the scenes? We trained a multimodal AI that links transcriptomes and text, enabling free-text search and annotation of RNA profiles. And we connected this model to an LLM that we fine-tuned into a chat assistant for transcriptome data (7/11)

🚀 We also validated CellWhisperer’s chat-based analysis with conventional bioinformatics. CellWhisperer was >4x faster (and 10x cooler 😊). Our recommendation: Use CellWhisperer for dataset exploration – but statistics is still important to ensure rigor & reproducibility (6/11)

🆕 The CellWhisperer paper (doi.org/10.1038/s415...) includes several new analyses beyond our 2024 bioRxiv preprint (biorxiv.org/content/10.1...). For example, we used CellWhisperer for an AI-guided analysis of human organ development (5/11)

🔬 You can easily query large transcriptome datasets for your favorite biological process using CellWhisperer. Just open Tabula Sapiens (cellwhisperer.cemm.at/tabulasapiens/) or GEO (cellwhisperer.cemm.at/geo/) in CellWhisperer & type your query into the chat box – for example “infection” (4/11)

🔍 We investigate one of the identified cell clusters by selecting the cells & prompting CellWhisperer with ‘Describe these cells in detail’. This interactive workflow is enabled by seamless integration of the CellWhisperer AI chat box into a version of CELLxGENE Explorer (3/11)

⚙️ To get started, let’s find cells by typing into the CellWhisperer chat box. For example ‘Show me structural cells with immune functions’. CellWhisperer scores each transcriptome by how well it matches this textual query and colors by query match (red: high, blue: low) (2/11)

🗨️Chat with your cells, explore #CellWhisperer!
CeMM PI Christoph Bock (@bocklab.bsky.social) together with @meduniwien.ac.at & @stanna-ccri.bsky.social have developed a new #AI tool that lets scientists explore single-cell data using plain English.

➡️ https://bit.ly/3JQqe87
📄 https://bit.ly/3WJCxGi

Systematic discovery of CRISPR-boosted CAR T cell immunotherapies - Nature CELLFIE, a CRISPR platform for optimizing cell-based immunotherapies, identifies gene knockouts that enhance CAR T cell efficacy using in vitro and in vivo screens.

📑 Check out our paper titled “Systematic discovery of CRISPR-boosted CAR T cell immunotherapies” at @Nature (open access): www.nature.com/articles/s41.... Feedback & suggestions are very welcome. (13/13)

🤝 This was a large project & great teamwork by: P. Datlinger*, E.V. Pankevich*, C.D. Arnold*, N. Pranckevicius, J. Lin, D. Romanovskaia, M. Schäfer, F. Piras, A.-C. Orts, A. Nemc, P. Biesaga, M. Chan, T. Neuwirth, A. Artemov, W. Li, S. Ladstätter, T. Krausgruber, C. Bock (12/13)

⚕️ Our CELLFIE platform supports clinical translation of CRISPR-boosted CAR T cells. For example, to avoid the DNA double-strand breaks introduced by CRISPR knockout, we performed a tiling base-editing screen across RHOG and identified promising gRNA for clinical testing. (11/13)

🔥 What’s next? Our discovery of strong combined effects for RHOG & FAS knockout underlines the potential of synergistic gene edits for boosting CAR T cell function. We thus integrated combinatorial screening into CELLFIE, using the Blainey lab’s CROPseq-multi method. (10/13)

🔬 From a technical perspective, we are excited how our new in vivo CROP-seq method improves gRNA detection (reading from an mRNA transcript as in nature.com/articles/nme...) and reduces experimental noise (by using UMIs), which enables larger screens with fewer mice. (9/13)

💪 We also observed prolonged survival for FAS knockout CAR T cells, likely because these cells are less effective at killing each other (“fratricide”). Combining RHOG & FAS knockout, we obtained more & better CAR T cells, which further improved survival in leukemic mice. (8/13)

🔍 RHOG is a small GTPase involved in cell signaling. How does it influence CAR T cells ? We found that RHOG knockout increases the proliferative capacity of CAR T cells and helps them retain a highly functional state with reduced exhaustion and enhanced memory phenotype. (7/13)