mjsushanth.github.io/protein-prediction.html at main · mjsushanth/mjsushanth.github.io · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>Protein Structure Prediction - Joel Markapudi</title>
    <style>
        * {
            margin: 0;
            padding: 0;
            box-sizing: border-box;
        }

        body {
            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, Oxygen, Ubuntu, Cantarell, sans-serif;
            line-height: 1.6;
            color: #333;
            background: #f5f5f0;
        }

        .container {
            max-width: 920px;
            margin: 0 auto;
            padding: 50px 30px 40px 30px;
            background: #fafaf5;
            min-height: 100vh;
        }

        .back-link {
            display: inline-flex;
            align-items: center;
            color: #4a90e2;
            text-decoration: none;
            margin-bottom: 30px;
            font-size: 0.95rem;
            transition: color 0.3s;
        }

        .back-link:hover {
            color: #357abd;
        }

        h1 {
            font-size: 2.2rem;
            font-weight: 400;
            margin-bottom: 18px;
            color: #2c2c2c;
            line-height: 1.3;
        }

        .project-overview {
            color: #555;
            font-size: 1.05rem;
            margin-bottom: 25px;
            padding-bottom: 20px;
            border-bottom: 2px solid #e0e0d8;
            line-height: 1.7;
        }

        .highlights {
            margin-top: 25px;
        }

        .highlights ul {
            list-style: none;
            margin: 0;
            padding: 0;
        }

        .highlights li {
            margin-bottom: 16px;
            padding-left: 20px;
            position: relative;
            color: #555;
            line-height: 1.7;
        }

        .highlights li:before {
            content: "•";
            position: absolute;
            left: 0;
            color: #4a90e2;
            font-weight: bold;
            font-size: 1.2rem;
        }

        strong {
            color: #2c2c2c;
            font-weight: 500;
        }

        .external-links {
            margin-top: 35px;
            padding-top: 25px;
            border-top: 1px solid #e0e0d8;
        }

        .external-links h2 {
            font-size: 1.3rem;
            font-weight: 500;
            margin-bottom: 15px;
            color: #2c2c2c;
        }

        .external-links a {
            display: inline-block;
            margin-right: 20px;
            margin-bottom: 10px;
            color: #4a90e2;
            text-decoration: none;
            font-size: 1rem;
            transition: color 0.3s;
        }

        .external-links a:hover {
            color: #357abd;
            text-decoration: underline;
        }

        @media (max-width: 768px) {
            .container {
                padding: 35px 20px;
            }

            h1 {
                font-size: 1.8rem;
            }
        }
    </style>
</head>
<body>
    <div class="container">
        <a href="index.html" class="back-link">← Back to Portfolio</a>

        <h1>Protein Structure Prediction</h1>

        <div class="project-overview">
            Implemented Hidden Markov Models, Conditional Random Fields, and BiLSTM architectures for secondary structure prediction on CB513 benchmark (514 non-homologous proteins). CRF model achieved 67% accuracy through specialized β-sheet detection features and position-specific evolutionary information.
        </div>

        <div class="highlights">
            <ul>
                <li><strong>Multi-Architecture Comparison:</strong> Built and evaluated HMM with mixture-of-Gaussians emissions, discriminative CRF with enhanced feature engineering, SVM with 533-dimensional PSSM-based features (74.91% accuracy), and BiLSTM with bidirectional context encoding</li>

                <li><strong>Feature Engineering:</strong> Designed 39,900-dimensional feature space incorporating PSSM evolutionary conservation scores, position-specific amino acid profiles, and 22 specialized β-sheet interaction features capturing N→N+3 residue dependencies</li>

                <li><strong>CRF Implementation:</strong> Achieved balanced three-state classification (helix/sheet/coil distributions: 0.364/0.289/0.347) through discriminative optimization with biologically meaningful structural transition modeling without explicit enforcement mechanisms</li>

                <li><strong>Architectural Analysis:</strong> Documented fundamental limitations of generative HMM approach (state collapse despite adaptive balancing) versus discriminative CRF success, demonstrating that long-range structural dependencies require feature-driven rather than probabilistic independence assumptions</li>

                <li><strong>Evaluation Framework:</strong> Comprehensive metrics including structure-specific F1 scores (SVM helix F1: 0.9522), confusion matrix analysis, and comparative study revealing strong correlation between evolutionary conservation patterns and structural predictability across 514-sequence benchmark</li>
            </ul>
        </div>

        <div class="external-links">
            <h2>Project Resources</h2>
            <a href="https://github.com/mjsushanth/ML_Protein_Structure_Prediction/blob/main/Report%20-%20Protein%20Struct%20Prediction%20HMM%2C%20CRF%2C%20LSTM%2C%20ML.pdf" target="_blank">→ Full Report (PDF)</a>
            <a href="https://github.com/mjsushanth/ML_Protein_Structure_Prediction" target="_blank">→ GitHub Repository</a>
        </div>
    </div>
</body>
</html>