You're reading the documentation for a development version. For the latest released version, please have a look at v0.1.5.

ContaminationDetector in action

In this example, we apply ContaminationDetector from coSMicQC on an example dataset from the NF1 project.

The NF1 project example includes wells from a cell line that was contaminated with mycoplasma. In the wet lab, these cells were detected as negative for mycoplasma. We do not want to process contaminated cells, so we can use this methodology to confirm the contamination and the extent of it on the plate.

The result of this method is either a pass or fail. If the data is clean, then the method stops at step 1 and says the data is ready for further downstream analysis. If the data has contamination, this method will continue processing after step 1 to determine if the problem is for the whole plate or part of the plate.

import pandas as pd

from cosmicqc import ContaminationDetector

# set a path for the NF1 parquet-based dataset
data_path = (
    "../../../tests/data/cytotable/NF1_cellpainting_data/Plate_3_filtered.parquet"
)

# Load in the dataset
filtered_nf1_df = pd.read_parquet(data_path)

# Look over the data to check it is correct
print(filtered_nf1_df.shape)
filtered_nf1_df.head()

(1355, 2321)

	Metadata_ImageNumber	Image_Metadata_Plate	Metadata_number_of_singlecells	Image_Metadata_Site	Image_Metadata_Well	Metadata_Cells_Number_Object_Number	Metadata_Cytoplasm_Parent_Cells	Metadata_Cytoplasm_Parent_Nuclei	Metadata_Nuclei_Number_Object_Number	Image_FileName_CY5	...	Nuclei_Texture_Variance_DAPI_3_02_256	Nuclei_Texture_Variance_DAPI_3_03_256	Nuclei_Texture_Variance_GFP_3_00_256	Nuclei_Texture_Variance_GFP_3_01_256	Nuclei_Texture_Variance_GFP_3_02_256	Nuclei_Texture_Variance_GFP_3_03_256	Nuclei_Texture_Variance_RFP_3_00_256	Nuclei_Texture_Variance_RFP_3_01_256	Nuclei_Texture_Variance_RFP_3_02_256	Nuclei_Texture_Variance_RFP_3_03_256
0	30	Plate_3	279	15	B11	1	1	2	2	B11_01_3_15_CY5_001_illumcorrect.tiff	...	619.327600	594.798669	271.137249	268.157417	311.088206	282.370923	198.402061	202.133683	203.094321	193.875072
1	31	Plate_3	279	16	B11	1	1	2	2	B11_01_3_16_CY5_001_illumcorrect.tiff	...	323.170295	321.310711	34.841145	35.139114	38.075206	38.080602	131.691809	126.174866	136.433036	132.735107
2	34	Plate_3	279	19	B11	1	1	2	2	B11_01_3_19_CY5_001_illumcorrect.tiff	...	321.457911	314.851226	286.810209	261.637391	257.878700	259.463388	157.252242	156.042241	154.576787	154.894240
3	35	Plate_3	279	1	B11	1	1	2	2	B11_01_3_1_CY5_001_illumcorrect.tiff	...	1487.354034	1468.971582	516.742751	489.945367	519.912829	510.173091	369.462002	366.631748	383.771987	364.529179
4	44	Plate_3	279	6	B11	1	1	2	2	B11_01_3_6_CY5_001_illumcorrect.tiff	...	508.054695	501.770497	51.695327	54.248623	57.984869	52.494053	262.420251	255.894670	259.081931	266.519397

5 rows × 2321 columns

# Instantiate the ContaminationDetector class and run the contamination detection process
detector = ContaminationDetector(
    dataframe=filtered_nf1_df, nucleus_channel_naming="DAPI"
).run()

Running step 1...
Summary:

Check                    Result
-----------------------  --------
Texture skewed?          ✅
Nucleus shape variable?  ❌
Interpretation:
Contamination detected! 🚨
Anomalous texture around nuclei detected but nuclei segmentation not clearly impacted.
Proceeding to step 2...
Running step 2...
Summary:

Check                                                    Result
-------------------------------------------------------  --------
Texture skewed?                                          ✅
Nucleus shape variable?                                  ❌
Whole plate contaminated due to abnormal texture?        ❌
Whole plate contaminated due to abnormal nucleus shape?  ❌
Interpretation:
Partial plate contamination in texture only. Proceed to step 3.
Running step 3...
Finding outlier cells with anomalous texture around the nucleus...
Number of outliers: 198 (14.61%)
Outliers Range:
Cytoplasm_Texture_InfoMeas1_DAPI_3_02_256 Min: -0.3808007336913458
Cytoplasm_Texture_InfoMeas1_DAPI_3_02_256 Max: -0.23152918999076016
Number of outliers: 120 (8.86%)
Outliers Range:
Cytoplasm_Granularity_2_DAPI Min: 1.8500394938002351
Cytoplasm_Granularity_2_DAPI Max: 27.639655282904474
Total number of outliers detected: 242

Number of wells in the top 25%: 3
Wells in the top 25% of highest outlier proportions:
B5, C5, D5

In this example, we can see that the detector has found anomalous texture surrounding the nucleus from this plate. This is a problem and could likely reflect contamination.

In step 2, based on the mean of the texture, it was found that this problem only impacts part of the plate.

In step 3, we found 3 wells that have high proportion of outlier single-cells with abnormal texture. This was concluded to be one cell line that had mycoplasma contamination on the plate, while the rest of the cell lines were fine.