{ "id": 463283, "title": "Computational method used for cell annotation", "url": "/competitions/open-problems-single-cell-perturbations/discussion/463283", "author_name": "", "post_date": "2023-12-24T11:15:36.020590800Z", "votes": 4, "comment_count": 1, "views": 0, "content": "

As discussed in this interesting post by @ambrosm https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661, a very important limitation is the cell type annotation.

\n

The populations of untreated and treated cells are not paired; we do not observe the same cell before and after treatment. Additionally, I believe that cell type annotation is more straightforward in the unperturbed state (e.g., in DMSO wells). The challenge arises when attempting to identify cell types in treatment wells. This issue is commonly encountered in these types of perturbational settings, which has led to the development of methods based on optimal transport, such as the one available at https://github.com/bunnech/cellot.

\n

What method was utilized for cell type annotation? For instance, having an estimation of uncertainty in cell type in the perturbed states could aid in the estimation of drug responses

", "messages": [ { "id": "2572573", "postDate": "12/24/2023 11:15:36", "content": "

As discussed in this interesting post by @ambrosm https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661, a very important limitation is the cell type annotation.

\n

The populations of untreated and treated cells are not paired; we do not observe the same cell before and after treatment. Additionally, I believe that cell type annotation is more straightforward in the unperturbed state (e.g., in DMSO wells). The challenge arises when attempting to identify cell types in treatment wells. This issue is commonly encountered in these types of perturbational settings, which has led to the development of methods based on optimal transport, such as the one available at https://github.com/bunnech/cellot.

\n

What method was utilized for cell type annotation? For instance, having an estimation of uncertainty in cell type in the perturbed states could aid in the estimation of drug responses

", "rawMarkdown": "As discussed in this interesting post by @ambrosm https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661, a very important limitation is the cell type annotation.\n\nThe populations of untreated and treated cells are not paired; we do not observe the same cell before and after treatment. Additionally, I believe that cell type annotation is more straightforward in the unperturbed state (e.g., in DMSO wells). The challenge arises when attempting to identify cell types in treatment wells. This issue is commonly encountered in these types of perturbational settings, which has led to the development of methods based on optimal transport, such as the one available at https://github.com/bunnech/cellot.\n\n**What method was utilized for cell type annotation?** For instance, having an estimation of uncertainty in cell type in the perturbed states could aid in the estimation of drug responses", "votes": null }, { "id": "2577157", "postDate": "12/28/2023 11:54:30", "content": "

Plus for the question.

\n

PS

\n

Just added some more umap plots to confirm AmbrosM conclusions that some cell types were misclassified:
\nhttps://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661#2576581

", "rawMarkdown": "Plus for the question. \n\nPS\n\nJust added some more umap plots to confirm AmbrosM conclusions that some cell types were misclassified:\nhttps://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661#2576581", "votes": null } ], "comments": [ { "id": 2577157, "author_name": "alexandervc", "author_url": "", "post_date": "12/28/2023 11:54:30", "content": "

Plus for the question.

\n

PS

\n

Just added some more umap plots to confirm AmbrosM conclusions that some cell types were misclassified:
\nhttps://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661#2576581

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "2572573": "As discussed in this interesting post by @ambrosm https://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661, a very important limitation is the cell type annotation.\n\nThe populations of untreated and treated cells are not paired; we do not observe the same cell before and after treatment. Additionally, I believe that cell type annotation is more straightforward in the unperturbed state (e.g., in DMSO wells). The challenge arises when attempting to identify cell types in treatment wells. This issue is commonly encountered in these types of perturbational settings, which has led to the development of methods based on optimal transport, such as the one available at https://github.com/bunnech/cellot.\n\n**What method was utilized for cell type annotation?** For instance, having an estimation of uncertainty in cell type in the perturbed states could aid in the estimation of drug responses", "2577157": "Plus for the question. \n\nPS\n\nJust added some more umap plots to confirm AmbrosM conclusions that some cell types were misclassified:\nhttps://www.kaggle.com/competitions/open-problems-single-cell-perturbations/discussion/458661#2576581" }, "source": "meta" }