Marcin Pacholczyk

Planning as Artificial Intelligence Problem - Short Introduction and Overview

Planning belongs to fundamental AI domains. Examples of planning applications are manufacturing, production planning, logistics and agentics. Over the decades planning techniques were improved and now they are able to capable real environment problems in the presence of uncertain and incomplete information. This article introduces the notion of so called classical planning, indicating connected with this computational complexity problems and possible ways of treating uncertainty.

Integrated System Supporting Research on Environment Related Cancers

There are many impediments to progress in cancer research. Insufficient or low quality data and computational tools that are dispersed among various sites are one of them. In this paper we present an integrated system that combines all stages of cancer studies, from gathering of clinical data, through elaborate patient questionnaires and bioinformatics tools, to data warehousing and preparation of analysis reports.

Exploring the Landscape of Protein-Ligand Interaction Energy Using Probabilistic Approach

Analysis of protein/small molecule interactions is crucial in the discovery of new drug candidates and lead structure optimization. Small biomolecules (ligands) are highly flexible and may adopt numerous conformations upon binding to the protein. Using computer simulations instead of sophisticated laboratory procedures may significantly reduce cost of some stages of drug development. Inspired by probabilistic path planning in robotics, stochastic roadmap methodology can be regarded as a very interesting approach to effective sampling of ligand conformational space around a protein molecule. Protein-ligand interactions are divided into two parts: electrostatics, modeled by the Poisson-Boltzmann equation, and van der Waals interactions, represented by the Lennard-Jones potential. The results are promising; it can be shown that locations of binding sites predicted by the simulation are in agreement with those revealed by experimental x-ray crystallography of protein-ligand complexes. We wanted to extend our knowledge beyond the current molecular modeling tools to arrive at a better understanding of the ligand-binding process. To this end, we investigated a two-level model of protein-ligand interaction and sampling of ligand conformational space covering the entire surface of protein target.

Probabilistic Approach to Planning Collision Free Path of UAV

In this chapter we present simulation results of an algorithm designed to planning collision free path based on probabilistic search. The aim of this study is to design an algorithm for off-line planning a set of way-points which make the collision free route of an Unmanned Aerial Vehicle (UAV). The planning process is based on the graph that represents the map containing information about altitude of the terrain over which the UAV is intended to perform its mission. The size of the graph increases polynomially with the resolution of the map and the number of way points of UAV. The probabilistic method of making the representative model of the terrain was applied in order to reduce a complexity of planning the collision free path. The functioning and efficiency of the approach proposed was introduced in our previous work, here we use real terrain map and introduce indicators to illustrate properties of the algorithm.

Influence of temperature on measurements of 3-axial accelerometers and gyroscopes: Embedded into inertial measurement unit

This paper presents the results of the study of influence of ambient temperature on the measurements of 3-axis linear acceleration sensors and gyroscopes used in IMU modules. The study was conducted using a climatic chamber, in which the six IMU modules were tested. Conducted tests covered a range of temperatures from -10 C to +60 C. The effect of ambient temperature on the values of zeros and gains of tested sensors was examined.

Different mutational characteristics of TSG in cell lines and surgical specimens

One of the most crucial concerns of cancer research pertains to the differences between the neoplastic cells in tumor specimens in vivo and their counterparts in cell lines. The huge amount of results deposited in cancer genetic databases allows to address this issue from a wider perspective. Our analysis of the Sanger Institute Catalog Of Somatic Mutations In Cancer (COSMIC) database v61 showed a lower percentage of homozygous mutations in a group of tumor suppressor genes in surgical samples (in vivo) in comparison to their frequency in cell lines (in vitro). Similarly, the mutations resulting in the lack of protein (e.g., nonsense mutations or whole gene deletions) of several tumor suppressor genes (TSGs) were more frequently observed in vitro than in vivo. In this article, we suggest two potential explanations of these data. Firstly, TSG heterozygous mutations resulting in the modified protein (e.g., missense mutations) may be gradually (when the specific molecular context is achieved) changed to homozygous mutations resulting in the lack of protein during carcinogenesis. Secondly, among different independent pathways of tumorigenesis, those leading to homozygous nonsense mutations are characteristic for cells which are more efficiently stabilized in vitro. To conclude, these observations may be interesting for researchers working with cell line in vitro models illustrating the extent to which they reflect the tumors in vivo.

Temperature Correction of Measurements Results of 3-Axis Accelerometers in IMU Modules

During temperature tests of 18 acceleration sensors in 6 IMU modules, conducted in a climate chamber at a temperature range from −10 to +60 °C, temperature characteristics of zeros and gains of tested sensors were obtained. An algorithm is presented for compensation of the sensor readings using temperature characteristics. We also report results of the effectiveness studies of temperature characteristics approximation by polynomials of various degrees.

Searching for Cancer Signatures Using Data Mining Techniques

Data mining finds many uses in biotechnology and one of them may be to analyze multi-platform data in order to allow searching for genomic cancer signatures. The importance of the topic arises as nowadays cancer is noted one of the leading causes of deaths in highly developed countries. The goal of this work was to search for colorectal cancer signatures, consisting of somatic mutations, somatic gene copy number alterations (SCNAs) as well as abnormal expression levels. After acquiring mutation, SCNA and expression data from cBioPortal, frequent itemset mining was performed using basket analysis and apriori algorithm. We also performed survival analysis of colorectal cancer patients using the discovered signatures as differentiating factor for Kaplan-Meier curve comparison. Frequent itemset mining returned modifications of genes that can be regarded as potential colorectal cancer signatures or signatures of carcinogenic processes in general. While methods used in the project consisted of use of simple or even basic tools, the results suggest that searching for cancer signatures amidst multi-platform data may be worth developing and improving.

EMQIT: a machine learning approach for energy based PWM matrix quality improvement

BackgroundTranscription factor binding affinities to DNA play a key role for the gene regulation. Learning the specificity of the mechanisms of binding TFs to DNA is important both to experimentalists and theoreticians. With the development of high-throughput methods such as, e.g., ChiP-seq the need to provide unbiased models of binding events has been made apparent. We present EMQIT a modification to the approach introduced by Alamanova et al. and later implemented as 3DTF server. We observed that tuning of Boltzmann factor weights, used for conversion of calculated energies to nucleotide probabilities, has a significant impact on the quality of the associated PWM matrix.ResultsConsequently, we proposed to use receiver operator characteristics curves and the 10-fold cross-validation to learn best weights using experimentally verified data from TRANSFAC database. We applied our method to data available for various TFs. We verified the efficiency of detecting TF binding sites by the 3DTF matrices improved with our technique using experimental data from the TRANSFAC database. The comparison showed a significant similarity and comparable performance between the improved and the experimental matrices (TRANSFAC). Improved 3DTF matrices achieved significantly higher AUC values than the original 3DTF matrices (at least by 0.1) and, at the same time, detected notably more experimentally verified TFBSs.ConclusionsThe resulting new improved PWM matrices for analyzed factors show similarity to TRANSFAC matrices. Matrices had comparable predictive capabilities. Moreover, improved PWMs achieve better results than matrices downloaded from 3DTF server. Presented approach is general and applicable to any energy-based matrices.EMQIT is available online at http://biosolvers.polsl.pl:3838/emqit.ReviewersThis article was reviewed by Oliviero Carugo, Marek Kimmel and István Simon.

Modeling Protein–Ligand Interaction with Finite Absorbing Markov Chain

We apply Stochastic Roadmap Simulation (SRS) and finite absorbing Markov chain theory to build a model of protein–ligand binding process. We evaluate a computational quantity—time to escape (TTE) from a funnel of attraction around binding site as a measure of binding affinity. The results based on PDBBind CoreSet (release 2008) show statistically significant correlation between experimental binding affinity and calculated TTE. Presented approach performs best for ligands with small number of internal degrees of freedom (rotatable bonds).