Skip to main content

Table 7 Feature importance calculated for the highest quality Spark supervised algorithms (Decision Trees (DT) and Random Forest (RF)). The entropy, the number of nodes that included certain features in the Random Forest building with RUS pre-processing and the average impurity decrease of the MLlib 2.0 Random Forest with ROS variants are presented for the alignment-based, alignment-free and alignment-based + alignment-free feature combinations The Random Oversampling pre-processing (ROS) is accompanied by the corresponding resampling size value

From: Surveying alignment-free features for Ortholog detection in related yeast proteomes by using supervised big data classifiers

  RUS + DT-Spark Weka RUS + RF-Spark/Gini Weka RF MLlib 2.0-Spark/Gini (Avg. Impurity Decrease)
Entropy Avg. Impurity Decrease Number of Nodes Normal ROS-100 ROS-130 RUS
Alignment-based Features/Algorithm
 nw 0.789 0.520 42 0.809 0.180 0.175 0.171
 sw 0.982 0.360 802 0.035 0.642 0.647 0.647
 profile3 0.783 0.360 417 0.043 0.167 0.167 0.167
 profile5 0.732 0.290 235 0.033 0.004 0.001 0.007
 profile7 0.712 0.240 330 0.080 0.008 0.010 0.008
Alignment-free Features
 aac 0.624 0.400 1891 0.033 0.173 0.171 0.169
 Auto_Geary 0.000 0.310 64 0.000 0.000 0.000 0.000
 Auto_Moran 0.000 0.320 75 0.000 0.000 0.000 0.000
 Auto_Total 0.000 0.370 1124 0.000 0.000 0.000 0.001
 CTD 0.408 0.310 1012 0.070 0.134 0.133 0.137
 CTD_C 0.566 0.300 1482 0.071 0.060 0.062 0.066
 CTD_D 0.407 0.320 1239 0.074 0.030 0.029 0.033
 CTD_T 0.529 0.290 1385 0.076 0.028 0.035 0.036
 fcm 0.265 0.310 1010 0.012 0.004 0.021 0.021
 2-mers 0.158 0.390 954 0.022 0.003 0.003 0.002
 2-mers_don’t care ps-1 0.000 0.320 847 0.000 0.000 0.000 0.000
 2-mers_ don’t care ps-2 0.000 0.310 768 0.001 0.000 0.000 0.000
 2-mers_ don’t care ps-3 0.000 0.260 772 0.000 0.000 0.000 0.001
 3-mers 0.078 0.370 1523 0.064 0.006 0.005 0.006
 3-mers_ don’t care ps-1 0.000 0.290 600 0.001 0.000 0.000 0.001
 3-mers_ don’t care ps-2 0.000 0.270 653 0.001 0.000 0.000 0.001
 3-mers_ don’t care ps-3 0.000 0.270 602 0.002 0.000 0.000 0.001
 length 0.507 0.400 2890 0.353 0.166 0.165 0.154
 nandy 0.109 0.260 902 0.009 0.000 0.000 0.001
 pseaa10 0.000 0.240 825 0.000 0.000 0.000 0.001
 pseaa3 0.611 0.380 1397 0.022 0.205 0.202 0.166
 pseaa4 0.609 0.380 1652 0.112 0.155 0.156 0.184
 QSO_maxlag_30_weight_01 0.280 0.240 1054 0.075 0.035 0.018 0.020
 QSOCN_maxlag_30 0 0.250 513 0.001 0.000 0.000 0.001
Alignment-based + Alignment-free Features/Algorithm
 nw 0.789 0.280 131 0.786 0.382 0.373 0.374
 sw 0.987 0.470 646 0.005 0.135 0.139 0.126
 profile3 0.769 0.280 271 0.005 0.098 0.101 0.097
 profile5 0.727 0.290 230 0.016 0.168 0.168 0.137
 profile7 0.710 0.260 229 0.004 0.083 0.084 0.126
 aac 0.623 0.190 230 0.015 0.073 0.071 0.072
 Auto_Geary 0.000 0.300 11 0.000 0.000 0.000 0.000
 Auto_Moran 0.000 0.270 11 0.000 0.000 0.000 0.000
 Auto_Total 0.000 0.510 147 0.001 0.000 0.000 0.000
 CTD 0.411 0.360 109 0.005 0.000 0.000 0.000
 CTD_C 0.570 0.340 204 0.039 0.032 0.032 0.032
 CTD_D 0.411 0.390 151 0.009 0.002 0.001 0.001
 CTD_T 0.531 0.320 164 0.001 0.002 0.003 0.004
 fcm 0.260 0.300 154 0.005 0.000 0.000 0.001
 2-mers 0.155 0.200 81 0.003 0.000 0.000 0.000
 2-mers_don’t care ps-1 0.000 0.410 104 0.000 0.000 0.000 0.000
 2-mers_ don’t care ps-2 0.000 0.410 98 0.000 0.000 0.000 0.000
 2-mers_ don’t care ps-3 0.000 0.400 82 0.001 0.000 0.000 0.000
 3-mers 0.074 0.230 97 0.010 0.000 0.000 0.000
 3-mers_ don’t care ps-1 0.000 0.390 69 0.000 0.000 0.000 0.000
 3-mers_ don’t care ps-2 0.000 0.340 49 0.001 0.000 0.000 0.000
 3-mers_ don’t care ps-3 0.000 0.390 59 0.001 0.000 0.000 0.000
 length 0.504 0.230 231 0.059 0.012 0.014 0.014
 nandy 0.113 0.320 101 0.001 0.000 0.000 0.001
 pseaa10 0.000 0.310 97 0.001 0.000 0.000 0.000
 pseaa3 0.613 0.190 142 0.009 0.006 0.007 0.004
 pseaa4 0.610 0.210 147 0.001 0.005 0.005 0.009
 QSO_maxlag_30_weight = 0.1 0.286 0.270 108 0.020 0.001 0.001 0.000
 QSO_maxlag_30 0.000 0.340 47 0.000 0.000 0.000 0.000
  1. nw: global alignment, sw: local alignment, profile: physicochemical profile from matching regions of aligned sequences at different window sizes (3, 5 and 7), aac: amino acid composition, pseacc: pseudo-amino acid composition at λ = 3,4 and 10, Auto_Geary: Geary’s auto correlation, Auto_Moran: Moran’s auto correlation, Auto_Total: Total auto correlation, fcm: four-color maps, nandy: Nandy’s descriptors, CTD: Composition, Distribution and Transition (Total), CTD_C: Composition, Distribution and Transition (Composition), CTD_D: Composition, Distribution and Transition (Distributions), CTD_T: Composition, Distribution and Transition (Transition), k-mers: 2-mers, 3-mers, spaced words: 2-mers with “don’t care positions” = 1, 2 and 3; 3-mer with “don’t care positions” = 1, 2, 3, QSO: Quasi-Sequence-Order, w = weight factor and maximum lag = 30
Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us