CartMFP 0.1.0

CartMFP: tools for database construction and formula prediction

CartMFP

CartMFP or Cartesian molecular formula prediction, is a python tool that can perform molecular formula predictions on custom databases.

How does CartMFP work

CartMFP consists of two steps.

1. Constructing a database space2cart.py : A local database is constructed.
   
2. Molecular formula prediction cart2form.py : Molecular formulas that are predicted based on input masses.
   
   

1. Constructing a database

A database is constructed by enumerating all combinations of elements within a certain range. The compositional space is described with a specific syntax: Element[min,max]. This can be any element in the periodic table for which the monoisotopic mass is described in the NIST database. "H[200]C[75]N[50]O[50]P[10]S[10]" is used default elemental space. This would eqaute to 0-200 Hydrogen, 0-75 Carbon, 0-50 Nitrogen, etc.

Apart from max element constraints, the elemental composition space is further limited by the maximum mass max_mass and ring double bond equivalents (RDBE) min_rdbe,max_rdbe.

Base chemical constraints:

Parameter	Default value	Description
-composition	"H[200]C[75]N[50]O[50]P[10]S[10]"	composition string describing minimum and maximum element counts
-max_mass	1000	maximum mass (Da)
-min_rdbe	-5	minimum RDBE
-max_rdbe	80	maximum RDBE

Additional chemical constraints are provided by implementing some of Fiehn's 7 Golden rules, which filters unrealistic or impossible compositions. This can drastically reduce the size of your composition space. These include: Rule #2 – LEWIS and SENIOR check; Rule #4 – Hydrogen/Carbon element ratio check; Rule #5 heteroatom ratio check and Rule #6 – element probability check.

Parameter	Default value	Description
-filt_7gr	True	Toggle global to apply or remove 7 golden rules filtering
-filt_LewisSenior	True	Golden Rule #2: Filter compositions with non integer dbe (based on max valence)
-filt_ratios	"HC[0.1,6]FC[0,6]ClC[0,2]BrC[0,2]NC[0,4]OC[0,3]PC[0,2]SC[0,3]SiC[0,1]"	#Golden Rules #4,5: Filter on chemical ratios with extended range 99.9% coverage
-filt_NOPS	True	#6 – element probability check.

Additional arguments can be supplied to affect the performance and output paths:

Parameter	Default value	Description
-maxmem	10e9	Amount of memory used in bytes
-mass_blowup	100000	blowup factor to convert float masses to integers
-write_mass	True	construct a mass lookup table (faster MFP but larger database)
-Cartesian_output_folder	"Cart_Output"	Path to output folder
-Cartesian_output_file		Output database name

Example use space2cart (In command line)

Contstruct default database:

python "space2cart.py" 

Contstruct database with halogens:

python "space2cart.py" -composition "H[200]C[75]N[50]O[50]P[10]S[10]F[5]Cl[5]I[3]Br[3]"

space2cart can also be executed by running the script in an IDE, such as Spyder.

2. Molecular formula prediction

After the composition database has been constructed with space2cart.py , molecular formula prediction can be done using cart2form.py. To run cart2form, an input mass list has to be supplied, which can be linked to a file in txt or any tabular format. The file either has to have a single column, or a column titled "Mz" or "Mass". Alternatively cart2form can be imported as a module within a script, and executed on a float mass or iterable set of masses with the function predict_formula(input_file,composition_file).

required inputs

Parameter	Default value	Description
-input_file	"test_mass_CASMI2022.txt"	path to list of masses
-composition_file	"H[200]C[75]N[50]O[50]P[10]S[10]_b100000max1000rdbe-5_80_7gr_comp.npy"	path to the database composition file

Optional arguments can supplied to tune which massses will be returned, this includes the polariy, which adducts to consider, which charge states to consider. Other key arguments include the ppm tolerance of the mass error of returned composition, and the maximum number of compositions returned per mass. Adducts use the following syntax: "sign(+ addition of /- loss of) elemental composition charge(+/-)"

Parameter	Default value	Description
-mode	"pos"	ionization mode. Options: " ", "positive", "negative"
-adducts	["+H+","+Na+","+K", "+-","+Cl-","-H+"]	default positive adducts "--","+H+","+Na+","+K", default negative adducts "+-","-H+","+Cl-"
-charges	[1]	Charge states to consider
-ppm	5	maximum mass error (ppm) of predicted compositions
-top_candidates	20	maxmimum number of compositions returned per mass

Example use cart2form

Molecular formula prediction from command line:

python "cart2form.py" -input_file "test_mass_CASMI2022.txt" -composition_file "H[200]C[75]N[50]O[50]P[10]S[10]_b100000max1000rdbe-5_80_7gr_comp.npy"

Alternatively cart2form.py can be imported.

import cart2form
cart2form.predict_formula(input_file=124.56 ,   #float mass or iterable (array/list/DataFrame)
                          composition_file="H[200]C[75]N[50]O[50]P[10]S[10]_b100000max1000rdbe-5_80_7gr_comp.npy")

Licensing

The pipeline is licensed with standard MIT-license.
If you would like to use this pipeline in your research, please cite the following papers:

Contact:

-Hugo Kleikamp (Developer): [e-mail]