exDEG: Python library for screening classification diagnostic gene
Where to get it
Binary installers for the latest released version are available at the pypi
#PyPI
#pip install exDEG
Dependencies
- Pyomic
See the full installation instructions for minimum supported versions of required, recommended and optional dependencies
Function available
| Functions | Description |
|---|---|
| get_hit_rate | Calculate the hit-rate of classifier |
| get_Qrecall | Calculate the Qrecall of classifier |
| get_avg_hit_rate | Calculate the average hit-rate of classifier |
| get_avg_Qrecall | Calculate the average Qrecall and PEM of classifier |
| cal_classgene | Calculate the performance of each gene as a classifier |
| find_exDEG | Find out genes that can be classified as diagnostic genes |
| find_classifyDEG | Find out the differential expression gene according to ROC/PEM and foldchange |
Algorithm principle (ROC, hit-rate, Qrecall)
An example is given to compare the effect of the classifier
Test data
Example
Input
import matplotlib.pyplot as plt
from sklearn import metrics
tmp_dict = {0.45:1,
0.34:0,
0.32:1,
0.26:1,
0.15:0,
0.14:0,
0.09:1,
0.07:0,
0.06:0,
0.03:0
}
total=10
hit_rate,avg_hit_rate = exDEG.get_hit_rate(tmp_dict,6,4)
Qrecall,avg_Qrecall,PEM = exDEG.get_Qrecall(tmp_dict,6,4)
fpr, tpr, thresholds=metrics.roc_curve(np.array(list(tmp_dict.values())),np.array(list(tmp_dict.keys())))
print()
print(avg_hit_rate,avg_Qrecall,PEM)
plt.plot(list(range(1,1+total)),hit_rate)
plt.plot(list(range(1,1+total)),Qrecall)
plt.legend(['hit_rate','Qrecall'])
plt.show()
roc_auc = metrics.auc(fpr, tpr)
#plt.subplot(grid[0,2:4])
plt.plot(fpr, tpr,'#EE1437')
plt.title('ROC curve (AUC = {1:0.2f})'
''.format(0, roc_auc))
plt.xlabel('Fpr')
plt.ylabel('Tpr')Output
Cal_classgene
Calculate the performance of each gene as a classifier
Receiver operating characteristic curve (ROC)
Input
class1=exDEG.cal_classgene(data,eg,cg,method='ROC',threshold=0.99)
class1.head()Output
| gene | auc | sig | |
|---|---|---|---|
| 0 | Fam175a | 0.416667 | normal |
| 1 | 1-Mar | 0.833333 | normal |
| 2 | Mob3b | 0.722222 | normal |
| 3 | Vapa | 0.583333 | normal |
| 4 | Crat | 0.083333 | normal |
Potential extract measure (PEM)
Input
class2=exDEG.cal_classgene(data,eg,cg,method='PEM',threshold=0.99)
class2.head()Output
| gene | pem | sig | |
|---|---|---|---|
| 0 | Fam175a | -0.166667 | normal |
| 1 | 1-Mar | 0.666667 | normal |
| 2 | Mob3b | 0.444444 | normal |
| 3 | Vapa | 0.166667 | normal |
| 4 | Crat | -0.833333 | normal |
find_classifyDEG
Calculate the performance of each gene as a classifier
Adjusting the threshold can change the number of genes screened
Receiver operating characteristic curve (ROC)
Input
#ROC result
res1=exDEG.find_classifyDEG(data,eg,cg,method='ROC',threshold=0.99,log2fc=-1)Output
...Calculated the class value of each gene
...Calculated the differential expression gene by classification
up: 40
down: 17
Potential extract measure (PEM)
Input
#PEM result
res2=exDEG.find_classifyDEG(data,eg,cg,method='PEM',threshold=0.99,log2fc=-1)Output
...Calculated the class value of each gene
...Calculated the differential expression gene by classification
up: 41
down: 23
