Changes between Dragon 3 and Dragon 5.1

Moreover, new molecular descriptors proposed by Estrada and derived from the edge adjacency matrix have been added together with some new molecular properties (Lipinski Alert Index, drug-like descriptors and three Base-Line Toxicities indices). The number of calculated molecular descriptors is thus increased up to 1630.

Since molecular walk counts can be very large numbers for big molecules, they were previously normalised by the factorial of the walk length. A better regularisation procedure is now achieved using a logarithmic transformation.

In Dragon 5.1 version descriptor calculations can be performed even on H-depleted molecule files: in this case, only H-invariant descriptors and 2D-autocorrelation descriptors are calculated.

The blocks of molecular descriptors have been augmented (from 18 to 20) and partially reorganized. More specifically:

▪	the added molecular descriptors proposed by Estrada can be found in the new block of edge adjacency indices.

▪	the aromaticity index block no longer exists, the four indices having been moved to the block of geometrical descriptors;

▪	the block of empirical descriptors no longer exists, the aromatic ratio (ARR) having been put into the constitutional descriptors and the unsaturation (UI) and hydrophylicity (Hy) indices into the block of molecular properties;

▪	the new blocks of connectivity, information and eigenvalue-based indices have been extracted from the previous block of topological descriptors;

▪	the new drug-like descriptors have been added in the molecular properties block:;

▪	finally, molecular path counts and related ID numbers, previously placed in the topological descriptor block, have been added to the molecular walk count block, thus creating the new block of walk and path counts.

In the frame 'Running the program' of Dragon's main window the button 'Load responses' has been added. This opens a new menu for importing user-defined variables (numerical and/or string). This menu has been conceived to import any ASCII file and enable the user to select the variables of interest. The selected variables will be loaded by Dragon and will be available, together with calculated molecular descriptors, for a preliminary correlation analysis and saving of the final output file until new variables are loaded by the user.

Principal components of the descriptor blocks can be selected, along with the descriptors of interest, in the 'Select descriptors' form in order to be available for graphical analysis of molecule distribution and correlation analysis with some user-defined variables before saving of the output file.

The standard Dragon output file has been reorganised so that first there are two fixed columns, one containing the sequential number of molecules (No.) and the other the molecule identifier (MolID) chosen by the user in the 'Select file' form. Then, the string variables, at most 3, eventually loaded by the user by the 'Load responses' menu, follow. Finally, the selected molecular descriptors, their PCs (if checked) and numerical user-defined variables (if loaded and selected) follow as before. The possibility of adding string variables in the output file is very important if additional information on molecules e.g., CAS number, product codes or specific names, needs to be stored.

In the 'Save descriptors' menu it is now possible to choose the code for missing values in the output file. Moreover, the skipping options have been partially changed to account for the modification of the Dragon output file format.

New interactive graph menus have been introduced in 'View descriptors'. Histogram graphs, Pareto plots and univariate statistics are available in the menu devoted to the analysis of each single descriptor. Line plots, 2D and 3D scatter plots are available in the new menu 'View molecules in the descriptor/response space'; these graphs allow a preliminary analysis of molecule distribution in the descriptor or principal component space, as well as a preliminary correlation analysis when user-defined responses have already been loaded.

In the 'View the most and least correlated descriptors with a selected one', the correlation thresholds can be defined by the user.

The history file (DragHist.log) has been implemented thus collecting all the information on the performed batch.