Changes from D6 to D7

Dragon 7 allows the calculation of molecular descriptors for disconnected

structures, such as salts. Dragon 7 offers new molecular descriptors (now a
total of 5270 divided into 30 logical blocks), hashed topological fingerprints,
molecular fragment analysis and new features for descriptor calculation and
input molecule management. Bugfixing and improvement of the algorithms
guarantee more efficient calculation performance and accuracy.

New features
New graphical user interface (GUI)
New molecular descriptors
Calculation of hashed fingerprint
Molecular fragment analysis
Calculation also for disconnected structures
Improved analysis tools. The tools for statistical analysis in the graphical
interface have been improved, adding the visualization of the correlation
map and the possibility of coloring mark points in 2D-scatterplots and PCA
plots
Weighting schemes. Atomic properties used as the weighting schemes for
descriptor calculation can be now easily accessed through the 'Periodic Table'
form (from the 'View' menu in the main menu bar)
New script wizard, template generation, and project recovery management

Bugfixing and algorithm improvement

Ring perception. The algorithm for the perception of rings in cyclic structures
has been revised and updated. This may lead to different values for some
descriptors (check the 'D7 descriptors with differences from the previous
version' section)
Faster implementation of correlation matrix calculation
Nitro-group standardization. The standardization procedure for nitro-groups
has been revised and updated. This may lead to different values for some
differences (check the 'D7 descriptors with differences from the previous
version' section)
Change in calculation of descriptors VR1_M, VR2_M, VR3_M in the block [7]
- 2D matrix-based descriptors (this may affect descriptor values of molecules
with multiple bond orders)
Bugfixed calculation of P_VSA-like descriptors; this could affect P_VSA like
values for molecules including atoms whose Van der Waals radii are not
defined and/or molecules with bonds whose reference bond lengths is not
defined (see the P_VSA-like descriptors page for more details). Bugfixed
introduced in Dragon version 7.0.4
Change in CATS 2D calculation; this affects CATS 2D descriptors at lag 0
(now, the algorithm follows the specifications as defined in Renner, S.,
Fechner, U., & Schneider, G. (2006). Alignment-free Pharmacophore Patterns
- A Correlation Vector Approach. In T. Langer & R. D. Hoffmann (Eds.),
Pharmacophores and Pharmacophore Searches (pp. 4979). Weinheim:
Wiley-VCH)
Verhaar baseline toxicities (block: [28] - Molecular properties) are not
calculated if MLogP < -6

For more details on changes in D7 descriptors click here.

New graphical user interface (GUI)

The layout for the graphical interface has been completely renewed, for a
better user experience. The main window has novel facilities for molecule
visualization and management. Input molecules are displayed in a scroll-down
sheet and can searched for by filtering tools. Clicking on a molecule, the
following facilities are activated:
- basic information on the chemical formula and composition
- visualization of the molecular structure
- atom weightings values
- molecular matrices from which molecular descriptors are calculated
- list of warning messages (if any)

New molecular descriptors

Block [1] - Constitutional descriptors (total number of new descriptors: 4): GD
(graph density); nTA (number of terminal atoms); nStructures (the number
of disconnected substructures), this descriptor is is related to the new
capability of Dragon of processing molecule with disconnected structures;
totalcharge (the sum of atomic formal charges)
Block [6] - Information indices (total number of new descriptors: 2): Ges
(Number of symmetry classes, based on electrotopological state); rGes
(Relative number of symmetry classes, based on electrotopological state)
Block [7] - 2D matrix-based descriptors (total number of new descriptors: 57):
VE1sign_M (coefficient sum of the last eigenvector from the matrix M),
VE2sign_M (average coefficient of the last eigenvector from the matrix M),
VE3sign_M (logarithmic coefficient sum of the last eigenvector from the
matrix M); Dragon implements 9 different graph matrices
Block [10] - P_VSA-like descriptors (total number of new descriptors: 10):
P_VSA_ppp_w (P_VSA-like on potential pharmacophore points), where w is
the property defining the potential pharmacophore points and can be: L lipophilic, P - positive charge, N - negative charge, D - hydrogen-bond donor, A
- hydrogen-bond acceptor, ar - aromatic, con - conjugated, hal - halogen, cyc belonging to cycles, ter - terminal
Block [14] - 3D matrix-based descriptors (total number of new descriptors: 9):
VE1sign_M (coefficient sum of the last eigenvector from the matrix M),
VE2sign_M (average coefficient of the last eigenvector from the matrix M),
VE3sign_M (logarithmic coefficient sum of the last eigenvector from the
matrix M); Dragon implements 3 different 3D molecular matrices
Block [23] - Atom-type E-state indices (total number of new descriptors: 2):
SaadN (Sum of aadN E-states); NaadN (Number of atoms of type aadN)
Block [29] - Molecular properties (total number of new descriptors: 1): Ro5
(Lipinski Rule of 5)
Block [30] - CATS 3D (total number of new descriptors: 300): this is the new
added descriptor block
For the complete list of the new molecular descriptors of Dragon 7 click here.

Disconnected structures
Disconnected structures, such as salts, can be processed by Dragon 7.

Different types of processing are available for molecular descriptor calculation

in the case of disconnected structures. Both molecular descriptors and
fingerprints can be calculated from disconnected structures. The descriptor
'nStructures' in the block [1] - Constitutional descriptors provides with the
count of the detected disconnected structures in each molecule.

Fingerprints and molecular fragments

Calculation of two types of hashed topological fingerprints has been
implemented in D7, namely Path FingerPrints (PFP) and Extended Connectivity
FingerPrints (ECFP). The user can choose among several different options in
order to customize the FP calculation. Moreover, the software generates the list
of fragments in the input molecule set; these fragments, codified by SMARTS
notation, can be visualized and exported in two different modes.