Why Cell-Type Resolution Matters in Drug Target Discovery
Most clinical trial failures trace back to targets identified from bulk tissue data. Cell-type resolution changes what discovery teams can see before a program begins.
Research notes
Thinking on single-cell genomics and drug discovery
Scientific writing from the Relation Therapeutics team: foundation models, single-cell methods, and the biology of target identification. Written by the people building the platform.
All articles
Foundation Models Meet Single-Cell RNA-seq: A New Paradigm
Transformer architectures trained on millions of cell profiles are beginning to do for biology what language models did for text. Here is what that means for target ID.
From Bulk Transcriptomics to Single-Cell: What Changes for Target ID
Bulk RNA-seq averages heterogeneous populations into a single signal. Switching to single-cell resolution changes not just sensitivity, but the shape of the hypotheses you can generate.
The Biology of Disease-Driving Cell Populations
Disease is not uniform across a tissue. Specific cell subtypes drive pathology while adjacent populations remain unaffected. Targeting the wrong population explains many program failures.
Handling Batch Effects in Large-Scale scRNA-seq Datasets
Batch effects are one of the biggest practical obstacles to training robust single-cell models across data sources. How we approach harmonization without erasing biological signal.
What Makes a Good Drug Target: Lessons from Failed Trials
Reviewing publicly documented late-stage failures reveals a pattern: many programs pursued targets with strong preclinical data but weak cell-type specificity in the relevant human tissue.
Embedding Single-Cell Data: Approaches and Tradeoffs
PCA, scVI, and transformer-based cell embeddings each make different assumptions about variance structure. Which one you choose shapes what the downstream model can learn.
Annotating Cell Types at Scale with Foundation Model Representations
Manual annotation of large scRNA-seq datasets is a bottleneck. Foundation model cell representations enable annotation transfer at scale while preserving rare subtype resolution.
The Role of Cell Atlases in Preclinical Target Validation
Human Cell Atlas and similar reference datasets give computational teams a shared biological map. How to use atlas reference data to contextualize your target hypothesis before animal work.
Building Training Datasets for Single-Cell Foundation Models
A foundation model is only as good as its training corpus. Curation decisions at the data-collection stage determine what biological variation the model can and cannot generalize from.
Integrating Multi-Modal Genomics for Target Hypothesis Generation
Combining scRNA-seq with ATAC-seq and spatial data gives a fuller picture of disease cell state. How multi-modal integration sharpens the target hypotheses that come out the other end.
Why Most Disease Models Miss the Cell-Type Signal
Mouse models, cell lines, and bulk RNA-seq each hide the cell-type dimension in different ways. Understanding why they fail is the first step toward tools that do not.