annadca 0.2.0

Annotation Assisted Direct Coupling Analysis

ANNotation Assisted Direct Coupling Analisis (annaDCA)

This package contains the methods and scripts to train and sample an RBM model provided with data annotations.

⬇️ Installation

Option 1: from PyPI

python -m pip install annadca

Option 2: cloning the repository

git clone https://github.com/rossetl/annaDCA.git
cd annaDCA
python -m pip install .

📘 Usage

After installation, all the main routines can be launched through the command-line interface using the command annadca. For now, only the train routine is implemented To see all the training options do

annadca train -h

✔️ Input data format

The training routine requires three types of information: the training sequences, their identifiers and the assocated annotations.

The package supports both binary variables and categorical variables (e.g. amino acid sequences).

These input data can be provided to the routine in two different ways:

• By providing a csv file to the -d argument that contains identifiers, sequences and annotations. By default, the routine will check for the columns name, sequence and label, but the user can specify different names using the arguments --column_names, --column_sequences and --column_labels.

[!WARNING] The previous method only works for categorical variables. For binary variables, use the following option

• By providing a fasta file (categorical variables) or a plain text file (binary variables) to the -d argument and an additional csv file to the -a argument containing the sequence identifiers and the annotations. Importantly, the sequence identifiers must match with those that are present in the fasta file.

To train the model with default arguments and a single csv file, use

annadca train -d <path_data> -o <output_directory> -l <model_tag>

Sequence data format

The model supports the following input data format:

• Binary variables: plain text format. Each row is one data sample, variables are separated by white spaces
• Categorical variables:
  • fasta format. Each data poin is a sequence of tokens with an header on top. The header row starts with >.
  • csv format: the file must contain one column with the aligned training sequences.