neuralbioinfo/PhaStyle-BACPHLIP

Dataset Card for PhageStyle-BACPHLIP

Dataset Summary:

The PhageStyle-BACPHLIP dataset was created to predict phage lifestyles, classifying them as either virulent or temperate. This dataset is split into training and validation sets, where the training set includes phages from various species (excluding Escherichia coli) and the validation set consists solely of Escherichia coli phages. The segmentation of sequences into 512bp and 1022bp fragments allows models to simulate real-world scenarios where metagenomic and viromic assemblies are often highly fragmented.

The dataset is particularly important for training machine learning models to generalize across different species and environments, providing robust predictions for phage lifestyle classification. This dataset was used to train the ProkBERT PhaStyle model.

Dataset Structure:

The dataset is segmented into 512bp and 1022bp fragments for both the training and validation sets:

Split Name	Number of Rows	Dataset Size (bytes)
Training_segmentdb_512	2,358,018	1,385,206,088
Training_segmentdb_1022	1,173,108	1,287,349,982
Validation_segmentdb_512	387,468	227,525,368
Validation_segmentdb_1022	192,745	211,472,614

• Training Segments: Contain phages from various species (excluding Escherichia coli) segmented into 512bp and 1022bp lengths. These fragments are used to train the model to predict the phage lifestyle, simulating short sequence fragments.
• Validation Segments: Consist exclusively of Escherichia coli phages and are segmented into 512bp and 1022bp fragments. These are used to validate the model's performance, especially its ability to generalize to Escherichia coli phages, a key bacterial species in many environments.

Dataset Splitting:

The dataset is split according to phage species, ensuring that Escherichia coli sequences are only present in the validation set. This ensures robust model generalization across other species during training.

---

Data Fields:

• segment_id: A unique identifier for each segmented entry.
• sequence_id: The ID of the original sequence from which the segment is derived.
• label: The lifestyle prediction for the phage (either virulent or temperate).
• y: Numeric encoding of the lifestyle label (1 for virulent, 0 for temperate).
• orientation: Indicates the orientation (forward or reverse-complement) of the sequence.
• Ls: The length of the segment in base pairs (512 or 1022).
• segment_start: The start position of the segment within the original sequence.
• segment_end: The end position of the segment within the original sequence.
• segment: The nucleotide sequence of the segmented portion.

---

Dataset Details:

The Training_segmentdb_512 split contains 512 base pair segments, with approximately 2.36 million rows, and is primarily used for training the phage lifestyle prediction models. The Training_segmentdb_1022 split contains 1022 base pair segments, with over 1.17 million rows. These larger segments provide more context for the model, potentially improving classification accuracy.

For model validation, the Validation_segmentdb_512 split includes 387,468 512bp segments, while the Validation_segmentdb_1022 split contains 192,745 1022bp segments. These validation splits allow for the evaluation of the model’s performance on both short and long fragments, mimicking fragmented phage sequences in real-world metagenomic assemblies. The validation set is exclusively composed of Escherichia coli phages, making it an important set for testing model generalization.

---

Dataset Structure:

The structure of the dataset is explained visually in the following figure:

Figure 1: The dataset used in the ProkBERT PhaStyle study. Phage sequences from multiple independent data sources were segmented into 512bp and 1022bp fragments for training and testing models on phage lifestyle prediction. The dataset consists of the BACPHLIP training and validation sets, Escherichia phages (from the Guelin collection), and phages from extreme environments.

---

Licensing

This dataset is provided under the same license as the ProkBERT PhaStyle repository. Please refer to the license for more details.

Citation

@Article{ProkBERT2024,
  author  = {Ligeti, Balázs and Szepesi-Nagy, István and Bodnár, Babett and Ligeti-Nagy, Noémi and Juhász, János},
  journal = {Frontiers in Microbiology},
  title   = {{ProkBERT} family: genomic language models for microbiome applications},
  year    = {2024},
  volume  = {14},
  URL={https://www.frontiersin.org/articles/10.3389/fmicb.2023.1331233},       
    DOI={10.3389/fmicb.2023.1331233},      
    ISSN={1664-302X}
}