Recent advances in sequencing allow the study of all identified human genes (≈ 22,000 protein encoding genes), which have differential expression between tissues. However, current knowledge on gene interactions lags behind, especially when one of the elements encodes a mitochondrial protein (≈ 1500). Mitochondrial proteins are encoded either by mitochondrial DNA (mtDNA; 13 proteins) or by nuclear DNA (nDNA; the remaining), which implies a coordinated communication between the two genomes. Since mitochondria coordinate several life-critical cellular activities, namely energy production and cell death, deregulation of this communication is implicated in many complex diseases such as neurodegenerative diseases, cancer and diabetes. Thus, this work aimed to identify high co-expression groups between mitochondrial genes-all genes, and associated protein networks in several human tissues (Genotype-Tissue Expression database). We developed a pipeline and a web tree viewer that is available at GitHub (https://github.com/Pereira-lab/CoExpression). Biologically, we confirmed the existence of highly correlated pairs of mitochondrial-all protein encoding genes, which act in pathways of functional importance such as energy production and metabolite synthesis, especially in brain tissues. The strongest correlation between mtDNA genes are with genes encoded by this genome, showing that correlation among genes encoded by the same genome is more efficient.

Cancer genome projects are characterizing the genome, epigenome and transcriptome of a large number of samples using the latest high-throughput sequencing assays. The generated data sets pose several challenges for traditional statistical and machine learning methods. In this work we are interested in the task of deriving the most informative genes from a cancer gene expression data set. For that goal we built denoising autoencoders (DAE) and stacked denoising autoencoders and we studied the influence of the input nodes on the final representation of the DAE. We have also compared these deep learning approaches with other existing approaches. Our study is divided into two main tasks. First, we built and compared the performance of several feature extraction methods as well as data sampling methods using classifiers that were able to distinguish the samples of thyroid cancer patients from samples of healthy persons. In the second task, we have investigated the possibility of building comprehensible descriptions of gene expression data by using Denoising Autoencoders and Stacked Denoising Autoencoders as feature extraction methods. After extracting information related to the description built by the network, namely the connection weights, we devised post-processing techniques to extract comprehensible and biologically meaningful descriptions out of the constructed models. We have been able to build high accuracy models to discriminate thyroid cancer from healthy patients but the extraction of comprehensible models is still very limited.

FP7 ICT4Depression project aims at providing a set of tools to further improve both patient outcome and increase of access to treatment of the patients suffering from major depression. This article describes the Information Systems (IS) architecture used in the project. ICT4Depression uses a service oriented architecture as means of bringing together different kinds of information concerning the patient, the therapeutic modules he is advised to follow and the sensors used to assess his status.

Abstract Inductive logic programming (ILP) is a sub‐field of machine learning that provides an excellent framework for multi‐relational data mining applications. The advantages of ILP have been successfully demonstrated in complex and relevant industrial and scientific problems. However, to produce valuable models, ILP systems often require long running times and large amounts of memory. In this paper we address fundamental issues that have direct impact on the efficiency of ILP systems. Namely, we discuss how improvements in the indexing mechanisms of an underlying logic programming system benefit ILP performance. Furthermore, we propose novel data structures to reduce memory requirements and we suggest a new lazy evaluation technique to search the hypothesis space more efficiently. These proposals have been implemented in the April ILP system and evaluated using several well‐known data sets. The results observed show significant improvements in running time without compromising the accuracy of the models generated. Indeed, the combined techniques achieve several order of magnitudes speedup in some data sets. Moreover, memory requirements are reduced in nearly half of the data sets. Copyright © 2008 John Wiley & Sons, Ltd.

One of the most well known successes of Inductive Logic Programming (ILP) is on Structure-Activity Relationship (SAR) problems. In such problems, ILP has proved several times to be capable of constructing expert comprehensible models that help to explain the activity of chemical compounds based on their structure and properties. However, despite its successes on SAR problems, ILP has severe scalability problems that prevent its application on larger datasets. In this paper we present LogCHEM, an ILP based tool for discriminative interactive mining of chemical fragments. LogCHEM tackles ILP's scalability issues in the context of SAR applications. We show that LogCHEM benefits from the flexibility of ILP, both by its ability to quickly extend the original mining model, and by its ability to interface with external tools. Furthermore, we demonstrate that LogCHEM can be used to mine effectively large chemoinformatics datasets, namely several datasets from EPA's DSSTox database and on a dataset based on the DTP AIDS anti-viral screen.

The amount of data collected and stored in databases is growing considerably for almost all areas of human activity. Processing this amount of data is very expensive, both humanly and computationally. This justifies the increased interest both on the automatic discovery of useful knowledge from databases, and on using parallel processing for this task. Multi relational data mining (MRDM) techniques, such as inductive logic programming (ILP), can learn rides from relational databases consisting of multiple tables. However, ILP systems are designed to run in main memory and can have long running times. We propose a pipelined data-parallel algorithm for ILP. The algorithm was implemented and evaluated on a commodity PC cluster with 8 processors. The results show that our algorithm yields excellent speedups, while preserving the quality of learning

ILP systems have been largely applied to data mining classification tasks with a considerable success. The use of ILP systems in regression tasks has been far less successful. Current systems have very limited numerical reasoning capabilities, which limits the application of ILP to discovery of functional relationships of numeric nature. This paper proposes improvements in numerical reasoning capabilities of ILP systems for dealing with regression tasks. It proposes the use of statistical-based techniques like model validation and model selection to improve noise handling and it introduces a search stopping criterium based on the PAC method to evaluate learning performance. We have found these extensions essential to improve on results over machine learning and statistical-based algorithms used in the empirical evaluation study.

The aim of behavioural cloning is to synthesize artificial controllers that are robust and comprehensible to human understanding. To attain the two objectives we propose the use of the Incremental Correction model that is based on a closed-loop control strategy to model the reactive aspects of human control skills. We have investigated the use of three different representations to encode the artificial controllers: univariate decision trees as induced by C4.5; multivariate decision and regression trees as induced by cart and; clausal theories induced by an Inductive Logic Programming (ILP) system. We obtained an increase in robustness and a lower complexity of the controllers when compared with results using other models. The controllers synthesized by cart revealed to be the most robust. The ILP system produced the simpler encodings.

A new model of human control skills is proposed and empirically evaluated. It is called the incremental correction model and is more adequate for reverse engineering human control skills than any other previously proposed models. The experimental results show a considerable increase in robustness of the controllers that use the new model. The new model also attenuates the problem of unbalanced classes, noticed already in previous experiments. By means of Parameterised Decision Trees, propositional learners are still usable within the new model's framework.