Omics Distillery
Tools probably useful for extracting mechanistic insight from omics data.
Modularised pipeline for understanding biological patterns (e.g. start with immune system)
-
ultimate goal is to reduce experiments to be done for drug discovery
-
when you capture a biological state in an experiment, what do you want to understand?
Model Zoo
Models and algorithms built for specific omics data analysis and interpretation purposes, that may be worth running in routine and complement each other in a flexible framework.
Geneformer
This is a Paragraph. Click on "Edit Text" or double click on the text box to start editing the content and make sure to add any relevant details or information that you want to share with your visitors.
scGPT
This is a Paragraph. Click on "Edit Text" or double click on the text box to start editing the content and make sure to add any relevant details or information that you want to share with your visitors.
Immune cell surface proteome interactome network
A physical wiring diagram for the human immune system
https://www.nature.com/articles/s41586-022-05028-x#Sec4
"From systems-level principles of immune cell connectivity down to mechanistic characterization of individual receptors, which could offer opportunities for therapeutic intervention."
​
1. Integrating our interactome with expression data to identify trends in the dynamics of immune interactions
2. A reductionist mathematical model that predicts cellular connectivity
3. Multi-tissue single-cell atlas that infers immune interactions throughout the body, revealing potential functional contexts for new interactions and hubs in multicellular networks
Deep Learning Biological Functional Representation of Gene Signatures (FRoGS)
Learn representations of “gene signatures projected onto their biological functions, instead of their identities”, “similar to word2vec”
“FRoGS vectors were trained such that individual human genes are mapped into high dimensional coordinates encoding their functions.”
“results in more effective compound-target predictions than models based on gene identities alone”
Drug target prediction through deep learning functional representation of gene signatures
Inferring gene regulatory networks by hypergraph variational autoencoder (incl gene encoder)
https://www.biorxiv.org/content/10.1101/2024.04.01.586509v1
"constructing GRNs, crucial to consider cellular heterogeneity and differential gene regulatory modules. However, traditional methods focused on cellular heterogeneity -> narrow scope"
"HyperG-VAE proves efficient in scRNA-seq data analysis and GRN inference."
Data Zoo
Non-exhaustive list of data that I would like to explore, build models with, and potentially use for transfer learning.
Human Commons Cell Atlas
Human Commons Cell Atlas contains uniformly preprocessed and filtered count matrices and cell type assignments that span 27 Human organs as well as marker gene lists for 31 Human organs.
CZ CellxGene Discover database
https://cellxgene.cziscience.com/datasets
​
Ensocell's preprint: "30 complex diseases in 13 disease-relevant tissues using the CZ CellxGene Discover database [21]."
Tool Zoo
Potentially useful tools that are not models on their own, such as methods that improves efficiency and lower computing costs.
Compute omics cheaper, scRNA and others, IntelLabs Open Omics
Inspiration Zoo
Discovery and insight from publications and posts.
Single-cell RNA sequencing of human tissue supports successful drug targets (Sarah Reichmann/ Ensocell)
https://www.medrxiv.org/content/10.1101/2024.04.04.24305313v1
"We estimate that combined they (cell type specific disease gene, disease cell overexpressed gene) could approximately triple the chances of a target reaching phase III."
"scRNA-seq support is more likely to prioritize therapeutically tractable classes of genes such as membrane-bound proteins."
"30 complex diseases in 13 disease-relevant tissues using the CZ CellxGene Discover database [21]."