Back

Unsupervised generative AI discovers pan-leukocyte dysregulated pathways in single-cell lupus data

Bolouri, H.; Cerosaletti, K.; Lacy-Hulbert, A.

2024-03-06 bioinformatics
10.1101/2024.03.02.583143 bioRxiv
Show abstract

Comparisons of single-cell RNA-sequencing (scRNA-seq) data from healthy and diseased tissues are widely used to identify cell-type-specific dysregulated genes, pathways, and processes. Accordingly, many sophisticated methods have been developed for this purpose. However, such tools generally require considerable user expertise for optimal performance. Here, we show that unsupervised application of a linearly-decoded Variational Auto Encoder (a generative AI model) to scRNA-seq data recapitulates and extends findings from a seminal recent lupus study and leads to new insights. Thanks to existing software libraries, our approach is straightforward to implement, computationally efficient, methodologically robust, and produces consistent and reproducible results.

Matching journals

The top 9 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.