bitser 0.5.21

Binary Pattern Sequence Recognition for biological sequences

BITSER

BInary paTtern SequencE Recognition

Overview

BITSER (Binary Pattern Sequence Recognition) is a software tool built with the Python language that extracts features segments of each genetic sequence at a local level.

The method for feature extraction utilizes the concept of Local Binary Pattern (LBP), as well as adapted versions of the Texture Unit and Texture Unit Number from the field of computer vision, to obtain informative features from sequences organized in FASTA files.

A k-mer window of size 9 slides over each genetic sequence, comparing the leftmost nucleotide or aminoacid in the window with the 8 other members.

This tool is targeted for usage by biologists, researchers and other professionals in the field of bioinformatics.

Installation

pip install bitser

After the installation, run bitser --help to see all the available commands.

CLI commands

BITSER follows a three-step workflow:

1. metadata → generate metadata.tsv with train/test split
2. train → extract features from train split and train model
3. predict → evaluate model on test split and generate reports

COMMAND	FUNCTION
metadata	Parse FASTA headers and create metadata.tsv with train/test splits
train	Extract features from training split and train classification model
predict	Load trained model, evaluate on test split, and generate reports

---

metadata command

Generates metadata.tsv by parsing FASTA headers and splitting data into train/test sets.

The dataset directory must contain a sequences/ subfolder with FASTA files.

Parameters

Parameter	Description	Required	Default
--dataset, -d	Dataset directory containing sequences/	✔
--class-delim, -delim	Delimiter used to extract class label from FASTA headers	✔
--train-count, -n	Number of sequences per class used for training	✔
--class-which, -which	Which occurrence of the delimiter to use (1 = first, -1 = last)		1
--seed	Random seed for reproducibility		7

Output

• metadata.tsv file containing dataset splits

---

train command

Performs feature extraction and trains a classification model using the training split only.

Parameters

Parameter	Description	Required	Default
--input, -i	Dataset directory containing metadata.tsv and sequences/	✔
--output-dir, -dir	Directory where outputs (model + logs) will be saved	✔
--output, -o	Model filename (e.g., model.pkl)	✔
--classifier, -c	Classifier: xgb, rf, svm, mlp, nb		xgb
--flank, -f	Sliding window size for feature extraction		8
--translate / --no-translate	Translate nucleotide sequences to proteins		False
--splits, -s	Number of cross-validation folds		10
--repeats, -r	Number of cross-validation repetitions		10
--seed	Random seed for reproducibility		7

Output

• Trained model (.pkl) saved inside --output-dir
• Training logs and evaluation results saved to --output-dir

---

predict command

Loads a trained model and evaluates it on the test split, generating predictions and reports.

Parameters

Parameter	Description	Required	Default
--model, -m	Path to trained model file	✔
--output-dir, -dir	Directory where prediction outputs will be saved	✔
--data, -d	Dataset directory containing metadata.tsv and sequences/	✔
--flank, -f	Sliding window size (must match training)		8
--translate / --no-translate	Must match training configuration		False

Output

• Classification results
• Per-class performance metrics
• Confusion matrix (if applicable)
• Prediction report (CSV) saved to --output-dir

Acknowledgements

• This study was supported by national funds through the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) - Finance Code 001, Fundação Araucária (Grant number 035/2019, 138/2021 and NAPI - Bioinformática), CNPq 440412/2022-6 and 408312/2023-8.