MutAIverse

The MutAIverse package streamlines adductomics research by enabling the screening of untargeted metabolomics data against a vast in silico MS/MS spectral library of Generative AI synthesized DNA adductome (MutAIverse spectral library) and experimentally validated adducts, facilitating the identification of DNA adducts that are formed due to genotoxin/carcinogen exposure along with the predictive capabilities of Adduct Linker module to determine potential source genotoxins responsible for the DNA adduct formation, establishing a direct link between DNA adducts and the genotoxins or carcinogens that trigger their formation. MutAIverse seeks to overcome the current limitations and pave the way for a more precise, efficient, and personalized assessment of the exposome.

The single strong dependency for this resource is RDKit, which can be installed in a local Conda environment.

Other dependencies

1. matchms (v0.13.0)
2. hnswlib (v0.8.0)
3. gensim (v4.3.3)
4. pandas (v1.5.3)
5. numpy (v1.23)
6. matplotlib (v3.7.1)
7. seaborn (v0.12.2)
8. tqdm (v4.65.0)

How to use MutAIverse?

Installation using pip

$ pip install -i https://test.pypi.org/simple/ MutAIverse

MutAIverse has 2 distinct modules:

1. Mapper
2. AdductLinker

Mapper module

Mapper module provides two approaches for mapping query MS spectra against in silico MS/MS spectral library of Experimentally validated adducts or Synthetic DNA adducts of MutAIverse.

MutAIverse Library setup

This is a one-time task that must be completed to use the Mapper module after installing the package for the first time.

from MutAIverse import Mapper
Mapper.load_library()

The function fetches the library data (1.5G) to be used by the Mapper module in the future.

Brute force Approach

Cosine Similarity-based mapping

from MutAIverse import Mapper
Mapper.map('bonafide_adducts',sample_file_path='/path-to-mzML-file',MS_level=1,plot=True)

Additional arguments

Parameters:
- library (str): bonafide_adducts/suspected_adducts/mutaiverse.
- sample_file_path (str): Path to the mzML file containing mass spectrometry data.
- ms level (int): 1 (MS spectrum) or 2 (MS/MS spectrum).
- plot (bool; default True): for visualizations.
return
- Result CSV file with suffix _MutAIversee_results.csv or _bonafide_adducts_results.csv.

Quick Search Approach

Approximate Nearest Neighbour-based mapping, which executes through 2 steps

1. Generation of spectral embeddings from query MS spectra
2. Mapping using the HNSW index of the spectral embeddings

from MutAIverse import Mapper
Mapper.fast_map(mzml_file_path)

Additional arguments

Parameters:
- mzml_file_path (str): Path to the mzML file containing mass spectrometry data.
- level (int; default 2): 1 (MS spectrum) or 2 (MS/MS spectrum)
- k (int; default 1): Number of nearest neighbors to search for.
- ef_query (int; default 300): Parameter controlling the number of elements to visit during a query.
- Energy (int; default 0): Collision energy

Returns:
- pandas.DataFrame: DataFrame containing search results with columns ['Query_Index', 'Nearest_Neighbor_Index', 'Cosine Similarity', 'SMILES', 'COMPID', 'Structures'].
- visualizations(density plot and histograms)

AdductLinker module

MutAIverse is also capable of re-tracing a DNA adduct to its possible source Genotoxin.

Fragment-based linking

Biotransformation backtracking based on abnormalities spliced from the base nucleotides.

from MutAIverse import Linker
query_smiles = 'OC[C@H]1O[C@H](CC1O)n1c[n+](c2c1nc(N)[nH]c2=O)C1OC2C(C1O)c1c(O2)cc(c2c1oc(=O)c1c2CCC1=O)OC' 
Linker.backtrace(Adduct = query_smiles)

Additional arguments

Parameters:
- Adduct (str): Path to the mzML file containing mass spectrometry data.
- knn (int; default 20): Number of nearest neighbors to narrow down the search space. 
- tophit (int; default 5): Minimum number of Genotoxins to be linked.
- plot (bool; default False): Traced SMILES 2D structures in rows.
- cutoff (int; default 80): Link Probability(%) cutoff.

Returns:
- pandas.DataFrame: DataFrame containing search results with columns ['Query', 'Fragment', 'Metabolites', 'N-Transformation', 'Genotoxin', 'Probability'].
- visualizations(Traced smiles 2D structures in rows)

This module also has a sub-function dedicated only to visualize backtrace() output with a user-supplied probability threshold.

import pandas as pd
from MutAIverse import Linker 
Linker.plot_trace(file='/Path-to-Output_file.csv')

Additional arguments

Parameters:
- file (str): Output CSV file (with path) of Linker.backtrace() function
- cutoff (int; default 80): Minimum probability threshold 
Returns:
- visualizations(Traced SMILES 2D structures in rows)