quasinet

Quasinet

Description

Infer non-local structural dependencies in genomic sequences. Genomic sequences are esentially compressed encodings of phenotypic information. This package provides a novel set of tools to extract long-range structural dependencies in genotypic data that define the phenotypic outcomes. The key capabilities implemented here are as follows:

1. Compute the Quasinet (Q-net) given a database of nucleic acid sequences. The Q-net is a family of conditional inference trees that capture the predictability of each nucleotide position given the rest of the genome. The constructed Q-net for COVID-19 and Influenza A H1N1 HA 2008-9 is shown below.

COVID-19	INFLUENZA

1. Compute a structure-aware evolution-adaptive notion of distance between genomes, which is demonstrably more biologically relevant compared to the standard edit distance.

2. Draw samples in-silico that have a high probability of being biologically correct. For example, given a database of Influenza sequences, we can generate a new genomic sequence that has a high probability of being a valid influenza sequence.

Installation

To install with pip:

pip install quasinet

To fix error with Mac or Windows:

from quasinet.osfix import osfix
# for windows
osfix('win')
# for max x86_64 (macbook pro)
osfix('macx86')
# mac arm (macbook air)
osfix('macarm')

NOTE: If trying to reproduce the paper below, please use pip install quasinet==0.0.58

Dependencies

• scikit-learn
• scipy
• numpy
• numba
• pandas
• joblib
• biopython

Usage

from quasinet import qnet

# initialize qnet
myqnet = qnet.Qnet()

# train the qnet
myqnet.fit(X)

# compute qdistance
qdist = qnet.qdistance(seq1, seq2, myqnet, myqnet) 

Examples

Examples are located here.

Documentation

For more documentation, see here.

Papers

For reference, please check out our paper:

Preparing For the Next Pandemic: Learning Wild Mutational Patterns At Scale For Analyzing Sequence Divergence In Novel Pathogens

Authors

You can reach the ZED lab at: zed.uchicago.edu