RT @jcheminf: We are looking for 2 new Associate Editors who share the vision for our journal to support the Editors-in-Chief in handling….

I shared some of my memories of my former mentor @atulbutte who has passed away.

RT @MSUMD: Congratulations to Teresa K. Woodruff, PhD, on receiving the National Medal of Science! Dr. Woodruff is the first MSU faculty me….

RT @aaguirrecano: The MSU Pharmacology & Toxicology has just posted a tenure-system Associate/Full Professor position for a visionary MSU D….

RT @ruairirobertson: A new bank of >2000 scientific/medical images and icons for free use in figures and presentations just released by @NI….

a great lineup of speakers. In silico Drug Discovery Workshop starting now

RT @labs_mann: Can a deadly rash be cured by repurposing drugs? Our @Nature paper says yes! Deep Visual Proteomics reveals JAK inhibitors a….

RT @TheMarkFdn: Open RFP: The Mark Foundation's Drug Discovery Award program is accepting letters of intent! This grant offers up to $1M in….

TransCell: In Silico Characterization of Genomic Landscape and Cellular Responses by Deep Transfer Learning . It took one year to formally publish it after acceptance.

RT @DrBinChen: 1/ Happy to share SPIDER is out @CellSystemsCP. Imputing abundance of over 2,500 surface proteins from single-cell transcri….

Thank you @NICHD_NIH for supporting two of our outstanding postdocs @michiganstateu. Proud to serve as the mentor for @rama_shankar12 's #K99 and as a co-mentor for Dr. Paul's #K99. This means a lot to them and me.

This project was led by my PhD student, Ruoqiao Chen, with significant contributions from Jiayu Zhou on model development. For those interested in SPIDER, please check out our R package (. We welcome your feedback!.

7/ To demonstrate the utility of the imputed data, we provided case studies in cell type annotation, biomarker/target identification, and cell-cell interaction in two cancers. In response to reviewers’ comments, we included 18 additional supplemental figures in the final version.

6/ Furthermore, only a few CITE-seq datasets were available, but we wanted to apply the model to datasets from various tissues, diseases, and cell types. Therefore, while developing SPIDER, we aimed to make the model context-agnostic and capable of supporting zero-shot prediction.

5/ We found a few prior studies that trained CITE-seq data for this purpose, but they were limited to the proteins profiled by CITE-seq, which typically includes only a few dozen proteins. However, there are over 5000 surface proteins.

4/ This work was also inspired by another project (, where we demonstrated that it is feasible to predict protein expression (and other genomic features) from gene expression.

3/ This work was motivated by our recent project (, from which we realized that most antibody-based therapies target surface proteins and it is crucial to consider protein expression in addition to transcript expression during target identification.

2/ Protein expression and transcript expression are not always correlated, and current experimental approaches can quantify only a few hundred proteins at most. Together with Prof. Jiayu Zhou, we present a model that could enable large-scale protein quantification.

1/ Happy to share SPIDER is out @CellSystemsCP. Imputing abundance of over 2,500 surface proteins from single-cell transcriptomes with context-agnostic zero-shot deep ensembles