From bd34dd02487ebd1f35fe7ba9babe59428e54cfc8 Mon Sep 17 00:00:00 2001 From: Hongzhi Wen Date: Tue, 13 Feb 2024 20:06:02 -0500 Subject: [PATCH] Update README.md --- README.md | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index bc4ba8f..f0ff4ad 100644 --- a/README.md +++ b/README.md @@ -25,9 +25,20 @@ pip install -r requirements.txt The full installation will install the same environment as we used during development. This includes `rapids` used to accelerate evaluation. ## Tutorials -We offer several [notebooks](https://github.com/OmicsML/CellPLM/tree/main/tutorials) for various downstream tasks as introductory tutorials. - -We are also working on developing more streamlined protocols for supported tasks and a comprehensive documentation. We aim to release these by the end of the year. +We offer several [notebooks](https://github.com/OmicsML/CellPLM/tree/main/tutorials) for various downstream tasks as introductory tutorials. _Our latest studies demonstrate CellPLM is competitive on cell-type annotation tasks compared to other SOTA methods and pretrained models. The result table is shown below:_ + +| Method | PBMC12K | Pancreas | HLCA | Immune | Brain | Liver | +| --- | --- | --- | --- | --- | --- | --- | +| SingleCellNet | 0.845+-0.0064 | 0.644+-0.0006 | 0.811+-0.0046 | 0.775+-0.0009 | 0.877+-0.0033 | 0.872+-0.0023 | +| ACTINN | 0.614+-0.0709 | 0.528+-0.0926 | 0.218+-0.0440 | 0.236+-0.0300 | 0.695+-0.0624 | 0.614+-0.0349 | +| scANVI | 0.930+-0.0148 | 0.963+-0.0083 | 0.708+-0.0183 | 0.851+-0.0133 | 0.933+-0.0010 | **0.908+-0.0144** | +| CellTypist | 0.883+-0.0055 | 0.882+-0.0011 | 0.776+-0.0079 | 0.822+-0.0020 | 0.901+-0.0031 | 0.764+-0.0132 | +| scDiff | 0.967+-0.0042 | **0.968+-0.0143** | **0.893+-0.0070** | 0.844+-0.0076 | 0.947+-0.0074 | 0.844+-0.0042 | +| scGPT | 0.963 | 0.954 | 0.863 | ***0.907*** | **0.950** | 0.864 | +| Geneformer | ***0.979*** | - | 0.833 | 0.856 | 0.934 | 0.871 | +| CellPLM | **0.975** | ***0.983*** | ***0.929*** | **0.902** | ***0.967*** | ***0.913*** | + +_(The evaluation follows the setting in [scDiff](https://www.biorxiv.org/content/10.1101/2023.10.13.562243v1.abstract) paper)_ ## Pretrained CellPLM Model Checkpoints The checkpoint can be acquired from our [dropbox](https://www.dropbox.com/scl/fo/i5rmxgtqzg7iykt2e9uqm/h?rlkey=o8hi0xads9ol07o48jdityzv1&dl=0). We might update our checkpoints from time to time.