Background Gene function annotations, which are associations between a gene and a term of a controlled vocabulary describing gene functional features, are of paramount importance in modern biology. approach, implementing two popular algorithms (Latent Semantic Indexing and Probabilistic Latent Semantic Analysis) and propose a novel method, the Semantic IMproved Latent Semantic Analysis, which adds a clustering step on the set of considered genes. Furthermore, we propose the improvement of these algorithms by weighting the annotations in the input set. Results We tested our methods and their weighted variants on the Gene Ontology annotation sets of three model organism genes (genes, the measure … has the same dimension of the original W matrix, Uis, the more confident the method is about the annotation to the feature term matrix. This property shows a limitation: on average, genes annotated to few terms tend to have a lower predicted annotation value in the be the matrix; given a gene annotation profile a, for each is computed as: tends to be low, and on average lower than the one obtained in the case when many values of a are not 0, i.e when a includes many annotations. In our tests, this was a clear source of bias when applying the tSVD predictive method to genes with a relevant buy Yunaconitine difference in the number of annotated terms. Because of this behavior, the predictive system using the tSVD approach tends to predict lot of annotations for well annotated genes and only a few for poorly annotated ones. Our Semantically IMproved tSVD (SIM) method is an attempt to overcome this issue, by adding a gene clustering step and defining a specific model for each cluster, i.e. group of more equally annotated genes. The V matrix of the tSVD algorithm implicitly uses the term-to-term correlation matrix T = WWthat approximates the input (weighted) annotation matrix W, pLSAnorm attempts to estimate the probability of the event can be interpreted as a multinomial probability distribution over the set of function terms and each entry of such vector is the probability of having a function term associated with the topic. Given the aspect model, the probability of an association between a gene are real kalinin-140kDa valued. Given a threshold … As an example of our gene annotation predictions, we report in Figure ?Figure55 a branch of the Directed Acyclic Graph of the GO Biological Process terms predicted by the SIM method, with the NTM weighting schema, as associated with the PGRP-LB Peptidoglycan recognition protein LB gene (Entrez Gene ID: 41379) of the Drosophila melanogaster organism. One may notice that, in this sub-tree, our SIM method predicted five new annotations, in addition to the six that were already present. Out of these five predicted annotations, two (catabolic process – GO:0009056 and macromolecole catabolic process – GO:0009057) were found validated with reliable evidence in the used dataset updated version. These confirmations suggest the likely correctness of their direct children, biopolymer catabolic process – GO:0043285 and carbohydrate catabolic process – GO:0016052, both also children of terms annotated to the same gene with reliable evidence in the dataset used for the prediction. Figure 5 Predictions for the PGRP-LB gene. Branch of the Directed Acyclic Graph of the GO Biological Process terms associated with the PGRP-LB buy Yunaconitine Peptidoglycan recognition protein LB gene (Entrez Gene ID: 41379) of the Drosophila melanogaster organism. It includes … Dataset version comparison buy Yunaconitine results In Table ?Table33 we report the validation results obtained by comparing the annotations predicted by each considered method and its weighting schema variants to the updated version of the annotation datasets used to generate the predictions. For each dataset, every prediction method returns a list of predicted annotations sorted according to their likelihood value. We considered the top 500 annotations of each list and evaluated the percentages of such annotations buy Yunaconitine that.