[2d CPCA block scores][2d CPCA block loadings][2d CPCA super weights][Grid CPCA block loadings][Active CPCA][Active CPCA differential][Active CPCA blocks]
The commands in this submenu starts different type of plot applications (2D plots, Grid plots). Follow these links for a full description of its options and commands.
The commands in the lower part of the menu are "active plots". They start a 2D plot application and as many plotGrid applications as field blocks contains the dataset. The 2D plot and the grid plots are intrinsically interconnected in such a way that the interaction of the User on the 2D plot will be immediately shown in all the grid plots.
The 2D application and the plotGrid application, when working together in an active plot, have essentially the same functionality than their stand-alone version, so please refer to the links in the previous paragraph for further reference regarding commands, options, etc...
Notice that this is a "tear-off menu", if you click on the dotted line on top of the menu you will transform the submenu in an independent window which will remain open until you specifically close it. This allows a fast access to the plots and is particularly useful when you plan to open many plots at the same time.
Plot>>>multiblock plots>>>2d CPCA block scores
Plot>>>multiblock plots>>>2d CPCA block loadings
First, the User is prompted to choose the component which will be represented in the X and Y axis as well as the block to represent.
When the dialog window shows the desired settings press the Plot button and the 2D plot will be displayed. The dialog will not close after producing the plot and will persists until the Exit button is pressed. This behavior allows to produce simultaneously the Loadings or Scores plot for more than one block without repeated access to the menus.
Block scores
Each block gives a different picture of the series, from the point of view of the block variables. The block scores are different for each block and different from the PCA scores.
The block scores can be seen as the scores of a "local" PCA that explain the variance within the block, but they are not equivalent to the scores that would have been obtained from a separate PCA made in each block. From the above algorithm, we can see that the scores were obtained using the block loadings (pb), which in turn were extracted from analysis of the overall model.
The block scores can be seen as the "opinion" of each block in the description of the series. This "opinion" is then elaborated at a higher level in order to obtain a "consensus" and therefore, some of the block description is lost and does not appear in the overall model.
When two objects/clusters appear close in the plot of block scores, the information in this block is not able to discriminate between them. If they are far from each other, the information in this block is able to discriminate them.
Block loadings
Are quite similar to the PCA loadings. They differ mainly in the scale, since they are normalized to length 1 for each block, while in PCA it is the whole loading vector which is normalized to length 1.
Plot>>>multiblock plots>>>2d CPCA super weights
First, the User is prompted to choose the components which will be represented in the X and Y.
When the dialog window shows the desired settings press the Plot button and the 2D plot will be displayed. The dialog will not close after producing the plot and will persists until the Exit button is pressed.
As mentioned in the background section, the superweights express the relative importance of each block in the overall model, for each model dimension.
The number and identity of the blocks more relevant for the overall model can be appreciated in the superweight plot. If some blocks appear together they are holding more or less the same information. Often this plot shows that one blocks participate in one PC and others blocks in a different PC.
Plot>>>multiblock plots>>>Grid CPCA block loadings
First, the User is prompted to choose the component which will be represented, as well as the block to show.
When the dialog window shows the desired settings press the Plot button and the grid plot will be displayed. The dialog will not close after producing the plot and will persists until the Exit button is pressed. This behavior allows to produce Loadings plots for more than one block without repeated access to the menus.
Block loadings are quite similar to the PCA loadings. They differ mainly in the scale, since they are normalized to length 1 for each block, while in PCA it is the whole loading vector which is normalized to length 1.
Plot>>>multiblock plots>>>Active CPCA
First, the User is prompted to choose the block to show.
When the right block was chosen press the OK button, or press the Cancel button to cancel.
The command will open:
The User can click with the mouse in any point of the block scores plot. A red cross is represented to mark the point. GOLPE extracts the values of the block scores for this point (t1 and t2) and projects back these values into the original space using the CPCA block loadings (P) to generate "pseudo-fields" which are represented in the grid plots.
The meaning of the pseudo-fields is equivalent to the meaning of the original fields used to obtain the CPCA; energies of interaction between the ligands and probes. The difference is that is obtained by a back-projection form a simplified model and therefore only the part of the variance explained by the two first PC's of the CPCA model is represented. Also, it is possible to obtain pseudo-fields for "theoretical compounds", not actually representing an object in the series.
For example: if the series contains two separate clusters of objects if is possible to click somewhere in between them to see how it would be like a compound sharing characteristics of the two clusters. Clicking in the middle of each clusters would produce a representation of the main characteristics which identify each cluster in the model, corresponding to the part of the variance explained by the two first PC's.
Active plots using this command are much similar to the ones which can be obtained using Plot>>>Active plots>>>PCA. However, the first are obtained using block scores instead of superscores and give a "more local" view of the data while the second show the results of the consensus for all the blocks.
Plot>>>multiblock plots>>>Active CPCA differential
First, the User is prompted to choose the block to show.
When the right block was chosen press the OK button, or press the Cancel button to cancel.
This command will open:
The User should click twice on the block scores plot. A red arrow is drawn between the first and the second clicked point (A and B points). GOLPE extracts the differences in the values of the block scores for this point (tB1-tA1 and tB2-tA2) and projects back these values into the original space using the PCA block loadings (P) to generate differences in "pseudo-fields" which are represented in the grid plots.
The meaning of these differences in pseudo-fields is equivalent to the meaning of differences of original fields computed for any two objects, as those that can be obtained using the command plot>>>Grid plots>>Objects differences. The difference is that is obtained by a back-projection form a simplified model and therefore only the part of the variance explained by the two first PC's of the CPCA model is represented. Also, it is possible to obtain differences in pseudo-fields for "theoretical compounds", not actually representing an couple of object in the series.
For example: if the series contains two separate clusters of objects if is possible to click in the middle of each clusters, to obtain a representation of the main characteristics which differentiate both clusters, corresponding to the part of the variance explained by the two first PC's. Drawing an arrow from a real compound or cluster of compounds to an empty region in the scores plot would represent the changes in the field required for the initial compounds to reach this region. A rational use of this plot can be very useful for the design of new compounds.
Active plots using this command are much similar to the ones which can be obtained using Plot>>>Active plots>>>PCA. However, the first are obtained using block scores instead of superscores and give a "more local" view of the data while the second show the results of the consensus for all the blocks.
Plot>>>multiblock plots>>>Active CPCA blocks
This command will open:
The User can click in two positions of the CPCA superscores plot (idem to the PCA scores plot) and the histogram will show the relative distance between the two clicked points in the different block scores plots. This plot is useful to understand the relevance of the different blocks in the discrimination between two objects or clusters of objects, and therefore is important in selectivity studies, mainly when more than two kinds of receptors are being studied.
The relative distances are computed as:
tD is the vector obtained as tB-tA, being tB and tA the points clicked by the User
from tD, a pseudofield X is computed
X = tD.p
And X can be spliced in different blocks X1…Xm
And from this formula we can obtain 
as
The length of the 
vector
is computed and normalized so the largest vector takes a value of 100.
Please notice that the vectors can be compared because the pb
are normalized to length 1, but still the values of tb
are dependent on the block scale. We recommend using this plot only on
data with consistent block scaling.