1 package jalview.datamodel;
3 import static org.testng.Assert.assertEquals;
5 import jalview.io.DataSourceType;
6 import jalview.io.FileParse;
7 import jalview.io.HMMFile;
8 import jalview.schemes.ResidueProperties;
10 import java.io.IOException;
11 import java.net.MalformedURLException;
14 import org.testng.annotations.BeforeClass;
15 import org.testng.annotations.Test;
17 public class HiddenMarkovModelTest {
19 HiddenMarkovModel hmm;
21 @BeforeClass(alwaysRun = true)
22 public void setUp() throws MalformedURLException, IOException
25 * load hmm model of a Kinase domain to a HiddenMarkovModel
27 HMMFile file = new HMMFile(new FileParse(
28 "test/jalview/io/test_PKinase_hmm.txt", DataSourceType.FILE));
32 @Test(groups = "Functional")
33 public void testGetMatchEmissionProbabilities()
34 throws MalformedURLException, IOException
37 * raw value in file is 3.67403
38 * saved as probability e^-X = 0.05259
40 double mep = hmm.getMatchEmissionProbability(0, 'R');
41 assertEquals(mep, 0.02537400637, 0.0001d);
42 assertEquals(mep, Math.pow(Math.E, -3.67403), 0.0001d);
44 mep = hmm.getMatchEmissionProbability(19, 'W');
45 assertEquals(mep, 0.00588228492, 0.0001d);
46 assertEquals(mep, Math.pow(Math.E, -5.13581), 0.0001d);
48 // column 160 is a gapped region of the model
49 mep = hmm.getMatchEmissionProbability(160, 'G');
50 assertEquals(mep, 0D, 0.0001d);
52 mep = hmm.getMatchEmissionProbability(475, 'A');
53 assertEquals(mep, 0.04995163708, 0.0001d);
54 assertEquals(mep, Math.pow(Math.E, -2.99670), 0.0001d);
57 @Test(groups = "Functional")
58 public void testGetInsertEmissionProbabilities()
60 double iep = hmm.getInsertEmissionProbability(2, 'A');
61 assertEquals(iep, Math.pow(Math.E, -2.68618), 0.0001d);
62 // symbol is not case-sensitive
63 assertEquals(iep, hmm.getInsertEmissionProbability(2, 'a'));
65 iep = hmm.getInsertEmissionProbability(5, 'T');
66 assertEquals(iep, Math.pow(Math.E, -2.77519), 0.0001d);
68 // column 161 is gapped in the hmm
69 iep = hmm.getInsertEmissionProbability(161, 'K');
70 assertEquals(iep, 0D, 0.0001d);
72 iep = hmm.getInsertEmissionProbability(472, 'L');
73 assertEquals(iep, Math.pow(Math.E, -2.69355), 0.0001d);
76 @Test(groups = "Functional")
77 public void testGetStateTransitionProbabilities()
79 // * in model file treated as negative infinity
80 double stp = hmm.getStateTransitionProbability(475,
81 HiddenMarkovModel.MATCHTODELETE);
82 assertEquals(stp, Double.NEGATIVE_INFINITY);
84 // file value is 5.01631, saved as e^-5.01631
85 stp = hmm.getStateTransitionProbability(8,
86 HiddenMarkovModel.MATCHTOINSERT);
87 assertEquals(stp, Math.pow(Math.E, -5.01631), 0.0001D);
89 stp = hmm.getStateTransitionProbability(36,
90 HiddenMarkovModel.MATCHTODELETE);
91 assertEquals(stp, Math.pow(Math.E, -5.73865), 0.0001D);
93 stp = hmm.getStateTransitionProbability(22,
94 HiddenMarkovModel.INSERTTOMATCH);
95 assertEquals(stp, Math.pow(Math.E, -0.61958), 0.0001D);
97 stp = hmm.getStateTransitionProbability(80,
98 HiddenMarkovModel.INSERTTOINSERT);
99 assertEquals(stp, Math.pow(Math.E, -0.77255), 0.0001D);
101 stp = hmm.getStateTransitionProbability(475,
102 HiddenMarkovModel.DELETETOMATCH);
103 assertEquals(stp, 1D, 0.0001D);
105 stp = hmm.getStateTransitionProbability(218,
106 HiddenMarkovModel.DELETETODELETE);
107 assertEquals(stp, Math.pow(Math.E, -0.95510), 0.0001D);
110 @Test(groups = "Functional")
111 public void testGetConsensusSequence()
113 SequenceI seq = hmm.getConsensusSequence();
114 String subStr = seq.getSequenceAsString().substring(0, 10);
115 assertEquals(subStr, "yelleklGsG");
116 subStr = seq.getSequenceAsString().substring(150, 161);
117 assertEquals(subStr, "-dllk------");
121 * A rather pointless test that reproduces the code implemented and asserts
122 * the result is the same...
124 @Test(groups = "Functional")
125 public void testGetInformationContent()
128 * information measure is sum over all symbols of
129 * emissionProb * log(emissionProb / background) / log(2)
131 Map<Character, Float> uniprotFreqs = ResidueProperties.backgroundFrequencies
135 for (char aa : hmm.getSymbols().toCharArray())
137 double mep = hmm.getMatchEmissionProbability(col, aa);
138 float background = uniprotFreqs.get(aa);
139 expected += mep * Math.log(mep / background);
141 expected /= Math.log(2D);
143 float actual = hmm.getInformationContent(col);
144 assertEquals(actual, expected, 0.0001d);