Proposal of supervised data analysis strategy of plasma miRNAs from hybridisation array data with an application to assess hemolysis-related deregulation

Landoni, Elena; Miceli, Rosalba; Callari, Maurizio; Tiberio, Paola; Appierto, Valentina; Angeloni, Valentina; Mariani, Luigi; Daidone, Maria Grazia

doi:10.1186/s12859-015-0820-9

Table 1 Model classification performance measures in the training and validation sets with raw and ratio data.

From: Proposal of supervised data analysis strategy of plasma miRNAs from hybridisation array data with an application to assess hemolysis-related deregulation

Training set								Validation set
Classication performance of the best performing groups of models					Parameters of the chosen model			Classication performance of the chosen model
Group ID	N models	Sens	Spec	Youden index	N miR	SVM cost	SVM weights	Sens [CI]	Spec [CI]	Youden index [CI]
1	16	0.85	0.96	0.81	35	10	(0.5; 0.5)	0.77 [0.54–0.92]	0.77 [0.61–0.92]	0.54 [0.23–0.81]
2	5	0.77	1.00	0.77	35	1	(0.5; 0.5)	0.85 [0.61–1.00]	0.81 [0.65–0.92]	0.65 [0.38–0.85]
3	2	0.77	0.96	0.73	30	1	(0.5; 0.5)	0.85 [0.61–1.00]	0.85 [0.69–0.96]	0.69 [0.42–0.88]
4	16	0.85	0.88	0.73	40	10	(0.5; 0.5)	0.77 [0.54–0.92]	0.73 [0.54–0.88	0.50 [0.19–0.77]
5	5	0.77	0.92	0.69	35	1	(0.4; 0.6)	0.85 [0.61–1.00]	0.73 [0.54–0.88]	0.58 [0.31–0.81]
6	16	0.85	0.85	0.69	50	10	(0.5; 0.5)	0.85 [0.61–1.00	0.69 [0.50–0.88	0.54 [0.27–0.81]
7	19	0.77	0.88	0.65	40	1	(0.4; 0.6)	0.77 [0.54–0.92]	0.69 [0.50–0.85]	0.46 [0.15–0.73]
8	5	0.69	0.92	0.61	3	100	(0.4; 0.6)	0.77 [0.54–0.92]	0.96 [0.88–1.00]	0.73 [0.46–0.92]
9	20	0.77	0.85	0.61	5	10	(0.4; 0.6)	0.69 [0.46–0.92]	0.92 [0.81–1.00]	0.61 [0.35–0.85]
Ratio data
1	1	0.92	0.81	0.73	500 (88)	0.01	(0.2; 0.8)	0.92 [0.77–1.00]	0.65 [0.46–0.85]	0.58 [0.31–0.81]
2	1	0.77	0.92	0.69	17 (16)	0.01	(0.3; 0.7)	0.77 [0.54–0.92]	0.92 [0.81–1.00]	0.69 [0.42–0.92]
3	1	0.69	1.00	0.69	90 (50)	0.01	(0.5; 0.5)	0.69 [0.38–0.92]	1.00 [1.00–1.00]	0.69 [0.38–0.92]
4	2	0.85	0.85	0.69	150 (66)	0.01	(0.2; 0.8)	0.92 [0.77–1.00]	0.69 [0.50–0.85]	0.61 [0.38–0.81]
5	35	0.69	0.96	0.65	4 (5)	0.1	(0.5; 0.5)	0.77 [0.54–0.92]	1.00 [1.00–1.00]	0.77 [0.54–0.92]
6	4	0.77	0.88	0.65	500 (88)	0.01	(0.4; 0.6)	0.92 [0.77–1.00]	0.77 [0.58–0.92]	0.69 [0.46–0.88]
7	3	0.85	0.81	0.65	600 (88)	0.01	(0.2; 0.8)	0.92 [0.77–1.00]	0.65 [0.46–0.85]	0.58 [0.31–0.81]
8	11	0.61	1.00	0.61	2 (3)	0.1	(0.5; 0.5)	0.77 [0.54–0.92]	1.00 [1.00–1.00]	0.77 [0.54–0.92]
9	23	0.69	0.92	0.61	3 (4)	0.1	(0.4; 0.6)	0.77 [0.54–0.92]	0.96 [0.88–1.00]	0.73 [0.50–0.92]
10	54	0.77	0.85	0.61	4 (5)	0.1	(0.3; 0.7)	0.77 [0.54–0.92]	0.88 [0.73–1.00]	0.65 [0.38–0.88]
11	18	0.61	0.96	0.58	3 (4)	0.1	(0.5; 0.5)	0.77 [0.54–0.92]	1.00 [1.00–1.00]	0.77 [0.54–0.92]
12	59	0.69	0.88	0.58	2 (3)	10	(0.4; 0.6)	0.77 [0.54–0.92]	1.00 [1.00–1.00]	0.77 [0.54–0.92]

In the last three columns, validation set classification performance measures are reported together with the corresponding bootstrap 95 % confidence intevals (CI)
Abbreviations: Group ID ID of the groups of best performing models (see also Fig. 2); N models number of models in each group, showing a specific classification performance, Sens sensitivity, Spec specificity, N miR number of miRNAs included in the model chosen in each group for containing the smallest number of miRNAs, SVM cost cost parameter of the linear SVM model, SVM weights weight parameter of the linear SVM model

Back to article page

ISSN: 1471-2105

Contact us

General enquiries: journalsubmissions@springernature.com

BMC Bioinformatics

Contact us