Classification of premalignant pancreatic cancer mass-spectrometry data using decision tree ensembles

BMC Bioinformatics

Table 1 Top ten features (m/z ratio) selected by Student t test method in our 10 fold cross validation.

Rank	Round 1	Round 2	Round 3	Round 4	Round 5	Round 6	Round 7	Round 8	Round 9	Round 10	Most Frequent
1	5798.9	5798.9	5819.8	5798.9	5819.8	5819.8	5798.9	5798.9	11477	5798.9	5798.9
2	5801.2	5819.8	5822.1	5801.2	5822.1	5822.1	5801.2	11541	11774	5801.2	5801.2
3	5819.8	5801.2	5798.9	5819.8	5798.9	5798.9	5819.8	11592	11472	11592	5819.8
4	5796.5	5822.1	5801.2	5822.1	5801.2	11592	5822.1	5801.2	5798.9	11597	11541
5	5822.1	11541	11592	5829.1	11770	11597	11541	11537	11481	11587	11592
6	11422	11592	11597	5831.4	11541	11587	5831.4	11546	5819.8	11541	5822.1
7	5817.4	11546	11541	11592	11597	5801.2	11592	11597	11770	11601	11597
8	11774	11587	11601	5803.5	11592	11541	11546	11774	11514	5819.8	11546
9	11541	11537	11546	11541	11601	11643	5829.1	11587	11509	11546	11601
10	11426	11569	11639	5796.5	11606	11601	11597	11601	5822.1	11606	11587

Rank is determined by the probability of the two means between disease and control groups in the training set being significantly different. m/z ratios with smaller probability ranks higher. Most frequent features are determined by the frequency of each feature appears in the top 10 list in these ten runs and ranked by their frequency.

ISSN: 1471-2105