P2-0209 — Annual report 2010
1.
Analysis of RNA-hnRNP C protein interactions using iCLIP data

We have developed algorithms and tools for mapping iCLIP sequence reads to the human genome, quality-control filtering, identification and removal of PCR duplicate reads and quantification of binding (cDNA counts) using data on random barcodes, generation of cross-link maps, identification of significant clusters of cross-links, and analysis of enriched pentanucleotides. Data analysis performed showed that hnRNP C recognizes uridine tracts with a defined long-range spacing (165nt) consistent with hnRNP particle organization.

COBISS.SI-ID: 7800916
2.
FragViz: visualization of fragmented networks

We have developed a new algorithm for computing layouts of networks consisting of a large number of disconnected components; such networks may appear naturally or by being intentionally broken into such components. The algorithm combines the Fruchterman-Reingold algorithms and multidimensional scaling with a new method for determining the optimal rotation of components. The proposed procedure was shown to be very useful for visualization of networks appearing in the field of bioinformatics.

COBISS.SI-ID: 7964756
3.
The differential ant-stigmergy algorithm

Ant-colony optimization is a stochastic algorithm for combinatorial problems. We proposed an upgrade capable of solving continuous numerical optimization problems. Such a problem is first transformed into a graph-search problem where graph vertices are labeled with differences of parameter values from currently best solution, and a computer-simulated population of ants is then used to iteratively search for paths in the graph representing better values of problem parameters. Experiments show the new algorithm to be particularly suitable for high-dimensional numerical problems.

COBISS.SI-ID: 23618855
4.
When is it better not to look ahead?

When solving search problems such as game-playing and path-finding, it is usually better to make decisions based on looking ahead for as many steps as possible. Search pathology occurs when looking ahead yields worse results. Key factors affecting the pathology are the number of heuristic values, the number of choices at each decision point, and the similarity between nearby choices. The effect of these factors is very similar in both minimax (games) and minimin search (path-finding and similar problems). The pathology was found in chess, the game of kalah and the eight-puzzle.

COBISS.SI-ID: 23912743
5.
High levels of uPA and PAI-1 predict a good response to anthracyclines

We demonstrated with the analysis of retrospective patient data the relation between markers uPA and PAI-1 and the response to various standard chemotherapies. High values of these markers suggest favorable response to the treatment with anthracyclines which is however more toxic. It is therefore important to predict the outcome of such treatment. The methodological novelty is the balancing of the compared subgroups. The paper was specially recommended in the same issue, with a review by a leading researcher in this field.

COBISS.SI-ID: 843131