Genomics and Machine Learning

DNA Sequence Analysis

One-Hot Encoding

import numpy as np

def one_hot_encode(sequence, seq_length=1000):
    nucleotides = {'A': 0, 'C': 1, 'G': 2, 'T': 3}
    encoding = np.zeros((seq_length, 4), dtype=np.float32)
    for i, base in enumerate(sequence[:seq_length]):
        if base in nucleotides:
            encoding[i, nucleotides[base]] = 1.0
    return encoding

Sequence Alignment with ML

Pairwise Alignment Score

Learned Scoring

class NeuralAligner(nn.Module):
    def __init__(self, embed_dim=32):
        super().__init__()
        self.nuc_embed = nn.Embedding(4, embed_dim)
        self.score_matrix = nn.Linear(embed_dim * 2, 1)

    def score_pair(self, nuc1, nuc2):
        e1, e2 = self.nuc_embed(nuc1), self.nuc_embed(nuc2)
        return self.score_matrix(torch.cat([e1, e2], dim=-1))

Variant Calling

Gene Expression Prediction

class ExpressionPredictor(nn.Module):
    def __init__(self, seq_length=10000):
        super().__init__()
        self.conv1 = nn.Conv1d(4, 64, kernel_size=15, padding=7)
        self.conv2 = nn.Conv1d(64, 128, kernel_size=15, padding=7)
        self.pool = nn.AdaptiveAvgPool1d(1)
        self.fc = nn.Sequential(nn.Linear(128, 64), nn.ReLU(),
                                nn.Dropout(0.3), nn.Linear(64, 1))

    def forward(self, x):
        x = torch.relu(self.conv1(x))
        x = torch.relu(self.conv2(x))
        return self.fc(self.pool(x).squeeze(-1))

Multi-Modal Genomic Models

Evaluation

• Pearson correlation for expression prediction
• AUC-ROC for variant pathogenicity
• Cohen's kappa: