Comments (16)
This can be implied that I should be aware not to overwrite metrics' results to runIntegration's results, right?
from scib.
from scib.
OK I got the point, so the result of Metrics.py is not correct in case of bkknn and cronos
from scib.
@mbuttner great that you spotted this! This is what we're trying to avoid with the unit tests. @mumichae can you take a look?
from scib.
@mbuttner Thanks for spotting it. I fixed it with 2cae16c on metric_fixes
.
@kridsadakorn Try, if it works now.
@LuckyMD I'm not quite sure how to set up unit tests that test the script. Should I wrap the decision part into a function of the package or should I call the script in the unit test?
from scib.
For now call the script in the test. We'll work on speed another time.
from scib.
Also, we should try to merge PRs a bit quicker. At the moment we're working in our own PRs too much. Let's make specific PRs and merge quickly.
from scib.
Could we merge the metrics_fixes right now already? Then I would open a new branch for the script testing
from scib.
let me take a look at them later/tomorrow. Just open a new branch for script testing and pull metric fixes into that.
from scib.
@kridsadakorn Has this issue been resolved?
from scib.
@mumichae How can I observe that the problem is gone? The ARI should not be equal to "1", right?
From Pancreas:
,human_pancreas_norm_bbknn_hvg0_knn,human_pancreas_norm_harmony_hvg0_embed,human_pancreas_norm_mnn_hvg0_full,human_pancreas_norm_scanorama_hvg0_embed,human_pancreas_norm_scanorama_hvg0_full,human_pancre
as_norm_seurat_hvg0_full,human_pancreas_norm_trvae_hvg0_embed,human_pancreas_norm_trvae_hvg0_full
NMI_cluster/label,0.46401801695452904,0.5042002919424581,0.6002282758722878,0.6123651119241046,0.6108255019101648,0.5679437049212369,0.5827635868924284,0.5388559433269677
ARI_cluster/label,0.16853642878982472,0.1514453414749276,0.20087193882907997,0.20956007009309266,0.20137486657273718,0.21995362399282753,0.20455721054365034,0.17647388431744748
ASW_label,,0.2852015053041666,0.509965711273253,0.42375880950884437,0.41808442771434784,0.3476724773645401,0.4333161190152168,0.273640975356102
ASW_label/batch,,0.32514855991659264,0.6220727460922302,0.5051653726518818,0.4626033526104369,0.4426532251891888,0.5113974121340542,0.3109629468094947
PCR_batch,,0.10250425238404949,0.06257392329011563,0.5766905597395535,0.5765105840018274,0.9603319486438531,0.39888335590252977,0.0166062273807448
cell_cycle_conservation,,0.8576605691022522,0.4196299205808169,0.3888340721973625,0.4194370296543745,0.6933786820742701,0.6689014648669532,0.9689979783426229
kBET,0.9811226904448863,0.3769181721402569,0.358476724390257,0.282121092915949,0.2860819109264171,0.63149884125723,0.3726876760945256,0.3726876760945256
iLISI,0.0,1.2251723136182102,1.1966802830580177,0.40344336386056434,0.40344336386056434,1.2357361497341808,0.4026809631922639,0.4026809631922639
cLISI,1.0,-0.0019265726299941832,-0.27309164785937856,0.6000196226157972,0.6000196226157972,0.37567982296272184,0.7273042624942341,0.7273042624942341
from scib.
@mumichae I updated for my group meeting today. People questioned that, for BBKNN, there are no values for ASW_label, PCR_batch, cell_cycle_conservation. Are they not applicable for BBKNN or error from the code? If they are not applicable, should it be better to have a value like NA or "-", at least we know that it is not an error.
from scib.
The overwriting of knn graphs is solved (2cae16c), so I'll close this issue. We should discuss the other problems in new issues.
from scib.
@mumichae I updated for my group meeting today. People questioned that, for BBKNN, there are no values for ASW_label, PCR_batch, cell_cycle_conservation. Are they not applicable for BBKNN or error from the code? If they are not applicable, should it be better to have a value like NA or "-", at least we know that it is not an error.
For these measures we use PCA, which is not available in knn-based methods, due to missing embedding or corrected expression value output. In other words, we can simply not compute these metrics on corrected kNN graphs.
from scib.
@mumichae How can I observe that the problem is gone? The ARI should not be equal to "1", right?
From Pancreas:
,human_pancreas_norm_bbknn_hvg0_knn,human_pancreas_norm_harmony_hvg0_embed,human_pancreas_norm_mnn_hvg0_full,human_pancreas_norm_scanorama_hvg0_embed,human_pancreas_norm_scanorama_hvg0_full,human_pancre as_norm_seurat_hvg0_full,human_pancreas_norm_trvae_hvg0_embed,human_pancreas_norm_trvae_hvg0_full NMI_cluster/label,0.46401801695452904,0.5042002919424581,0.6002282758722878,0.6123651119241046,0.6108255019101648,0.5679437049212369,0.5827635868924284,0.5388559433269677 ARI_cluster/label,0.16853642878982472,0.1514453414749276,0.20087193882907997,0.20956007009309266,0.20137486657273718,0.21995362399282753,0.20455721054365034,0.17647388431744748 ASW_label,,0.2852015053041666,0.509965711273253,0.42375880950884437,0.41808442771434784,0.3476724773645401,0.4333161190152168,0.273640975356102 ASW_label/batch,,0.32514855991659264,0.6220727460922302,0.5051653726518818,0.4626033526104369,0.4426532251891888,0.5113974121340542,0.3109629468094947 PCR_batch,,0.10250425238404949,0.06257392329011563,0.5766905597395535,0.5765105840018274,0.9603319486438531,0.39888335590252977,0.0166062273807448 cell_cycle_conservation,,0.8576605691022522,0.4196299205808169,0.3888340721973625,0.4194370296543745,0.6933786820742701,0.6689014648669532,0.9689979783426229 kBET,0.9811226904448863,0.3769181721402569,0.358476724390257,0.282121092915949,0.2860819109264171,0.63149884125723,0.3726876760945256,0.3726876760945256 iLISI,0.0,1.2251723136182102,1.1966802830580177,0.40344336386056434,0.40344336386056434,1.2357361497341808,0.4026809631922639,0.4026809631922639 cLISI,1.0,-0.0019265726299941832,-0.27309164785937856,0.6000196226157972,0.6000196226157972,0.37567982296272184,0.7273042624942341,0.7273042624942341
I don't quite understand what you mean. The ARI scores are between 0 and 1. They are indeed quite small, which means that the cluster labels and cell type labels don't match very well according to ARI
from scib.
from scib.
Related Issues (20)
- SCANVI bug
- Graph connectivities metric HOT 1
- Low values for iLISI score
- error when running fast metrics HOT 2
- Normalize: could not find function "sizeFactors" HOT 1
- Import scib error "but is an incompatible architecture (have 'x86_64', need 'arm64e" MacOXS Monterrey 12.5.1 with Apple M1 ultra HOT 2
- scib.me.ilisi_graph: FileNotFoundError: [Errno 2] No such file or directory: '/tmp/lisi_svo3el2i/graph_lisi_indices_0.txt' HOT 5
- Does scgen uses layers.counts or adata.X normalized counts, and where is this indicated? HOT 2
- Missing kBET argument documentation
- ilisi scores HOT 1
- Rpy2 as optional dependency
- Running SCIB HOT 3
- Cannot understand the mean of scib.metrics.pcr_comparison parameter HOT 1
- kBET module not found error is supplanted by an np.nan return HOT 2
- `cell_cycle` returns poor scores on perfect data input HOT 4
- GPU support for scib metrics? HOT 2
- Not really able to import scib HOT 2
- Preprocessing best practices, scvi wrapper usage, anndata inputs to scib.metrics
- ilisi_graph error HOT 3
- AttributeError: Can only use .cat accessor with a 'category' dtype (when running lisi() metric) HOT 3
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from scib.