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

Table 5 Percentage of true positives (%TP) identified by both outstanding supervised and unsupervised classifiers when detecting ortholog pairs in the twilight zone (< 30% of identity)

Algorithm/Dataset	Alignment-based Features			Alignment-free Features			Alignment-based + Alignment-free Features
Algorithm/Dataset	%TP			%TP			%TP
Supervised Algorithms	Scer Cgla	Cgla Klac	Klac Kwal	Scer Cgla	Cgla Klac	Klac Kwal	Scer Cgla	Cgla Klac	Klac Kwal
Spark Random Forest MLlib 1.6
Normal	0.00	0.00	0.00	0.00	0.00	0.00	3.54	0.00	2.25
ROS-100	97.43	96.26	98.03	71.06	64.29	57.87	96.14	91.84	93.54
ROS-130	98.71	96.94	98.31	76.21	64.97	65.45	95.18	93.88	93.54
RUS	99.04	96.26	98.60	74.60	64.29	61.24	96.78	93.88	95.51
Spark Decision Trees MLlib 1.6
Normal	0.32	0.68	0.28	0.00	0.00	0.56	12.54	7.82	9.55
ROS-100	95.18	94.56	97.19	72.67	62.93	55.62	97.75	84.69	96.07
ROS-130	95.82	91.50	97.47	79.74	61.56	63.48	98.71	87.41	96.35
RUS	98.07	95.24	99.16	76.53	67.01	65.45	98.07	90.82	97.47
Unsupervised Algorithms
RBH	57.56	58.84	73.31
RSD 0.2 1e-20	46.95	45.92	62.36
RSD 0.5 1e-10	61.41	61.90	80.34
RSD 0.8 1e-05	68.17	70.41	85.96
OMA	42.77	45.24	46.91

The best results in each dataset are in bold face and the general best results are underlined. Supervised algorithm performance is presented for the alignment-based, alignment-free and alignment-based + alignment-free feature combinations

ISSN: 1471-2105