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);
63 iep = hmm.getInsertEmissionProbability(5, 'T');
64 assertEquals(iep, Math.pow(Math.E, -2.77519), 0.0001d);
66 // column 161 is gapped in the hmm
67 iep = hmm.getInsertEmissionProbability(161, 'K');
68 assertEquals(iep, 0D, 0.0001d);
70 iep = hmm.getInsertEmissionProbability(472, 'L');
71 assertEquals(iep, Math.pow(Math.E, -2.69355), 0.0001d);
74 @Test(groups = "Functional")
75 public void testGetStateTransitionProbabilities()
77 // * in model file treated as negative infinity
78 double stp = hmm.getStateTransitionProbability(475,
79 HiddenMarkovModel.MATCHTODELETE);
80 assertEquals(stp, Double.NEGATIVE_INFINITY);
82 // file value is 5.01631, saved as e^-5.01631
83 stp = hmm.getStateTransitionProbability(8,
84 HiddenMarkovModel.MATCHTOINSERT);
85 assertEquals(stp, Math.pow(Math.E, -5.01631), 0.0001D);
87 stp = hmm.getStateTransitionProbability(36,
88 HiddenMarkovModel.MATCHTODELETE);
89 assertEquals(stp, Math.pow(Math.E, -5.73865), 0.0001D);
91 stp = hmm.getStateTransitionProbability(22,
92 HiddenMarkovModel.INSERTTOMATCH);
93 assertEquals(stp, Math.pow(Math.E, -0.61958), 0.0001D);
95 stp = hmm.getStateTransitionProbability(80,
96 HiddenMarkovModel.INSERTTOINSERT);
97 assertEquals(stp, Math.pow(Math.E, -0.77255), 0.0001D);
99 stp = hmm.getStateTransitionProbability(475,
100 HiddenMarkovModel.DELETETOMATCH);
101 assertEquals(stp, 1D, 0.0001D);
103 stp = hmm.getStateTransitionProbability(218,
104 HiddenMarkovModel.DELETETODELETE);
105 assertEquals(stp, Math.pow(Math.E, -0.95510), 0.0001D);
108 @Test(groups = "Functional")
109 public void testGetConsensusAtAlignColumn()
111 assertEquals(hmm.getConsensusAtAlignColumn(10), 's');
112 assertEquals(hmm.getConsensusAtAlignColumn(50), 'k');
113 hmm.setConsensusResidueStatus(false);
114 assertEquals(hmm.getConsensusAtAlignColumn(100), 'l');
115 assertEquals(hmm.getConsensusAtAlignColumn(400), 'k');
118 @Test(groups = "Functional")
119 public void testGetConsensusSequence()
121 SequenceI seq = hmm.getConsensusSequence();
122 String subStr = seq.getSequenceAsString().substring(0, 10);
123 assertEquals(subStr, "YELLEKLGSG");
124 subStr = seq.getSequenceAsString().substring(150, 161);
125 assertEquals(subStr, "-DLLK------");
129 * A rather pointless test that reproduces the code implemented and asserts
130 * the result is the same...
132 @Test(groups = "Functional")
133 public void testGetInformationContent()
136 * information measure is sum over all symbols of
137 * emissionProb * log(emissionProb / background) / log(2)
139 Map<Character, Float> uniprotFreqs = ResidueProperties.backgroundFrequencies
143 for (char aa : hmm.getSymbols())
145 double mep = hmm.getMatchEmissionProbability(col, aa);
146 float background = uniprotFreqs.get(aa);
147 expected += mep * Math.log(mep / background);
149 expected /= Math.log(2D);
151 float actual = hmm.getInformationContent(col);
152 assertEquals(actual, expected, 0.0001d);