Toward mastering the cell language by learning to generate

Chen, Yixin; Bian, Haiyang; Wei, Lei; Jia, Jinmeng; Dong, Xiaomin; Li, Yuzhuo; Zhao, Yubo; Wu, Xinze; Li, Chen; Luo, Erpai; Xiao, Chuxi; Hao, Minsheng; Zhang, Xuegong

doi:10.1101/2024.01.25.577152

Cited by 3 publications

(1 citation statement)

References 78 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this way, we can have cPeaks as the equivalence of the transcriptome in scRNA-seq processing, which will make comparison and integration of data from different studies uniform and standardized. Currently, the emergence of various foundation models based on transcriptions has provided a significant understanding of life science [74][75][76][77] . It's also essential to establish foundation models for chromatin accessibility data.…”

Section: Discussionmentioning

confidence: 99%

Toward a generic feature set defined by consensus peaks as a consistent reference for ATAC-seq data

Meng

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

Chromatin accessibility profiling methods such as assay for transposase-accessible chromatin using sequencing (ATAC-seq) have been promoting the identification of gene regulatory elements and the characterization of epigenetic landscapes. Unlike gene expression data, there is no consistent reference for chromatin accessibility data, which hinders large-scale integration analysis. By analyzing many more than 1000 ATAC-seq samples and 100 scATAC-seq samples, we found that cells share the same set of potential open regions. We thus proposed a reference called consensus peaks (cPeaks) to represent open regions across different cell types, and developed a deep-learning model to predict all potential open regions in the human genome. We showed that cPeaks can be regarded as a new set of epigenomic elements in the human genome, and using cPeaks can increase the performance of cell annotations and facilitate the discovery of rare cell types. cPeaks also performed well in analyzing dynamic biological processes and diseases. cPeaks can serve as a general reference for epigenetic studies, much like the reference genome for genomic studies, making the research faster, more accurate, and more scalable.

show abstract

Section: Discussionmentioning

confidence: 99%

Toward a generic feature set defined by consensus peaks as a consistent reference for ATAC-seq data

Meng

et al. 2023

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

Profiling cell identity and tissue architecture with single-cell and spatial transcriptomics

Gulati,

D’Silva,

Liu

et al. 2024

Nat Rev Mol Cell Biol

View full text Add to dashboard Cite

scGenePT: Is language all you need for modeling single-cell perturbations?

Istrate,

Li,

Karaletsos

2024

Preprint

View full text Add to dashboard Cite

Modeling single-cell perturbations is a crucial task in the field of single-cell biology. Predicting the effect of up or down gene regulation or drug treatment on the gene expression profile of a cell can open avenues in understanding biological mechanisms and potentially treating disease. Most foundation models for single-cell biology learn from scRNA-seq counts, using experimental data as a modality to generate gene representations. Similarly, the scientific literature holds a plethora of information that can be used in generating gene representations using a different modality - language - as the basis. In this work, we study the effect of using both language and experimental data in modeling genes for perturbation prediction. We show that textual representations of genes provide additive and complementary value to gene representations learned from experimental data alone in predicting perturbation outcomes for single-cell data. We find that textual representations alone are not as powerful as biologically learned gene representations, but can serve as useful prior information. We show that different types of scientific knowledge represented as language induces different types of prior knowledge. For example, in the datasets we study, subcellular location helps the most for predicting the effect of single-gene perturbations, and protein information helps the most for modeling perturbation effects of interactions of combinations of genes. We validate our findings by extending the popular scGPT model, a foundation model trained on scRNA-seq counts, to incorporate language embeddings at the gene level. We start with NCBI gene card and UniProt protein summaries from the genePT approach and add gene function annotations from the Gene Ontology (GO). We name our model "scGenePT", representing the combination of ideas from these two models. Our work sheds light on the value of integrating multiple sources of knowledge in modeling single-cell data, highlighting the effect of language in enhancing biological representations learned from experimental data.

show abstract

Toward mastering the cell language by learning to generate

Cited by 3 publications

References 78 publications

Toward a generic feature set defined by consensus peaks as a consistent reference for ATAC-seq data

Toward a generic feature set defined by consensus peaks as a consistent reference for ATAC-seq data

Profiling cell identity and tissue architecture with single-cell and spatial transcriptomics

scGenePT: Is language all you need for modeling single-cell perturbations?

Contact Info

Product

Resources

About