Assessing Outlier Probabilities in Transcriptomics Data When Evaluating a Classifier

. 2023 Feb 1;14(2):387. doi: 10.3390/genes14020387

Algorithm 1 Pseudo-code of bootstrap–bagplot algorithm for outlier detection for two study groups.

1:
Initialize lists to report the samples that are randomly drawn from the original dataset D in the bootstrap runs and the outliers that are detected in each case:

$S a m p l e s = []$

$O u t l i e r s = []$
2:
for $i = 1, 2, \dots, I$ do
3:
Create a bootstrap dataset $B_{i}$ by resampling the original dataset D with replacement.
4:
Report the sample IDs present in $B_{i}$ , i.e., $S a m p l e s_{B_{i}}$ with

$S a m p l e s = a p p e n d (S a m p l e s, S a m p l e s_{B_{i}})$ .
5:
Conduct a principal component analysis for $B_{i}$ including samples from both study groups.
6:
Split $B_{i}$ into the two study group subsets $B_{i 1}$ and $B_{i 2}$ .
7:
for $k = 1, 2$ do
8:
Detect outliers of k-th study group subset $B_{i k}$ using bagplot:

$O u t l i e r s_{B_{i k}} = b a g p l o t (B_{i k}, f a c t o r = 2)$
9:
Report the samples being detected as outliers with

$O u t l i e r s = a p p e n d (O u t l i e r s, O u t l i e r s_{B_{i k}})$
10:
end for
11:
Aggregate $S a m p l e s$ list to frequency table.
12:
Aggregate $O u t l i e r s$ list to frequency table.
13:
Compute relative outlier frequencies and conduct binomial test with $α = 0.05$ .
14:
end for