The chosen solution based on local descriptors is later evaluated in a cross-database scenario with the three mentioned datasets, and full dataset 5-fold cross validation.
The mentioned datasets are challenging for image classification because of their large intra class variance and inter class relationship.
The categories used in our annotation scheme are a union of the actions employed in the three above mentioned datasets, i.e., (i) general facial actions, (ii) facial actions with object manipulation, (iii) general body movements, (iv) body movements with object interaction, and (v) body movements for human interaction.
Spectral normalisation is achieved in all of the mentioned datasets.
It is also worth pointing out that the three mentioned datasets are quite representative of microarray cancer studies.
Since all of the mentioned datasets have been widely used in previous studies, we provide only a synoptic description of each of them in the Supplementary File, where the interested reader can find relevant references for a more in-depth description of them.
More specifically, we repeated each alignment in the previously mentioned datasets by using (i) the normal procedure, enabling the program to build its own guide-tree, (ii) forcing the use of the correct tree (or a canonical species tree) as a guide-tree and (iii) forcing the use as a guide-tree of a synthetic 'wrong' tree having the maximum split-distance to the correct tree.
We employed the ROC analysis to compare our algorithm to other prior methods for their ability to identify the TFBS motifs of structural and/or evolutionarily related TFs in the above mentioned three datasets, for Dataset-1 and Dataset-2, the ROC curves were plotted based on the following criteria.
(i) Realistic dataset with variable length (RealVar dataset): This dataset consists of total 14725 BCEs (i.e., 11981 IgG-specific, 2341 IgE-specific, 403 IgA-specific BCEs) and 22835 non-BCEs of variable lengths (Table 4) (ii) Balanced dataset with variable length (BalanceVar dataset): This dataset was created from the above mentioned RealVar dataset.
