RNA embedding

RNA Embedding Model

This tutorial will guide you through how to use the RNA embedding model using the OmniGenomeModelForEmbedding class. We will cover initializing the model, encoding RNA sequences, saving/loading embeddings, and computing similarities.

Install Required Dependencies

Before we start, make sure you have the necessary libraries installed. You can install them using the following command:

!pip install OmniGenome torch transformers autocuda

Setting Up the Embedding Model

First, let's initialize the OmniGenomeModelForEmbedding class with a pre-trained model.

from omnigenome import OmniGenomeModelForEmbedding

# Initialize the model using a pre-trained model path (replace with RNA-specific model if available)
model_name = "bert-base-uncased"  # Example model, replace with your own model
embedding_model = OmniGenomeModelForEmbedding(model_name)

Encoding RNA Sequences into Embeddings

We'll now encode a batch of RNA sequences into embeddings.

# Example RNA sequences (replace these with your own RNA sequences)
rna_sequences = [
    "AUGGCUACG",
    "CGGAUACGGC",
    "UGGCCAAGUC",
    "AUGCUGCUAUGCUA"
]

# Encode the RNA sequences into embeddings
rna_embeddings = embedding_model.batch_encode(rna_sequences)

# Display the generated embeddings
print("RNA Embeddings:")
print(rna_embeddings)

Saving and Loading Embeddings

You can save the generated embeddings to a file and load them later when needed.

# Save embeddings to a file
embedding_model.save_embeddings(rna_embeddings, "rna_embeddings.pt")

# Load the embeddings from the file
loaded_embeddings = embedding_model.load_embeddings("rna_embeddings.pt")

# Display the loaded embeddings to verify
print("Loaded RNA Embeddings:")
print(loaded_embeddings)

Computing Similarity Between RNA Sequences

Let's compute the similarity between two RNA sequence embeddings using cosine similarity.

# Compute the similarity between the first two RNA sequence embeddings
similarity = embedding_model.compute_similarity(loaded_embeddings[0], loaded_embeddings[1])

# Display the similarity score
print(f"Similarity between the first two RNA sequences: {similarity:.4f}")

Encoding a Single RNA Sequence

You can also encode a single RNA sequence into its embedding.

# Example single RNA sequence
single_rna_sequence = "AUGGCUACG"

# Get the embedding for the single RNA sequence
single_rna_embedding = embedding_model.encode_single_sequence(single_rna_sequence)

# Display the embedding for the single RNA sequence
print("Single RNA Sequence Embedding:")
print(single_rna_embedding)