Dependencies¶
ALFABET¶
ALFABET is a graph neural network model for predicting bond dissociation energies and
free energies of non-cyclic bonds in organic molecules. It was trained on a large dataset of
DFT-calculated data obtained at the M06-2X/def2-TZVP level of theory. The training dataset contained
neutral closed-shell molecules composed of C, H, N, and O. [1]
DBSTEP¶
DBSTEP is a software package for calculating steric descriptors from 3D molecular
structures. It can be used through BONAFIDE to calculate the buried volume and buried shell volume
descriptors. [2]
DScribe¶
DScribe is a Python library that converts 3D atomic or molecular structures into fixed-length
numerical fingerprints, e.g., as input for machine learning models.
[3] [4] Within BONAFIDE, DScribe was used to implement Atom-centered
Symmetry Functions, Smooth Overlap of Atomic Positions, Local Many-body Tensor Representation, and
an atomic Coulomb Matrix-based descriptor. [5]
kallisto¶
kallisto is a command-line tool for rapidly calculating various atomic features from 3D
molecular structures, that is, coordination number, partial charge, (relative)
polarizability, the proximity shell descriptor, and the van der Waals radius.
[6]
mendeleev¶
mendeleev is a Python package that provides access to an extensive database of the properties
of the chemical elements (e.g., various atomic radii or electronegativity scales), which can be
accessed through BONAFIDE. [7]
MORFEUS¶
MORFEUS is a software package that provides a broad range of molecular as well as atom and bond
descriptors calculated from 3D molecular structures. BONAFIDE implements MORFEUS’ buried volume,
dispersion, local force constant, and solvent-accessible surface area as well as several
angle-based atom and bond descriptors, respectively. [8]
Multiwfn¶
Multiwfn is a multifunctional stand-alone program for the analysis of the electronic structure
of a molecule after a quantum chemical calculation. [9]
[10] It provides access to a large number of atom- and bond-level features
that can be calculated through BONAFIDE. For an overview of the available features, see
List of features.
Psi4¶
Psi4 is an open-source quantum chemistry software package that is used in BONAFIDE to perform
single-point energy calculations - e.g., with density functional theory (DFT), including
dispersion corrections and implicit solvation models. [11]
[12] [13]
qmdesc¶
qmdesc is a graph neural network model for predicting atom and bond property descriptors that
are typically obtained through quantum chemical calculations. This includes Hirshfeld partial
charges, Fukui indices, NMR shielding constants, as well as bond orders and
lengths. The large training dataset contained neutral closed-shell molecules composed of C, H,
O, N, P, S, F, Cl, Br, and I and was obtained at the B3LYP/def2-SVP level of theory.
[14]
RDKit¶
RDKit is the general cheminformatics package used in BONAFIDE. It also implements a number of
atom and bond descriptors which can be calculated through BONAFIDE. For the full list, see the
List of features. [15]
xtb¶
xtb is a semi-empirical quantum chemistry software package that is used in BONAFIDE to perform
single-point energy calculations as well as for the calculation of various atom features
such as reactivity indices from conceptual DFT (C-DFT). [16]
References