Logo: to the web site of Uppsala University

In pharmaceutical research there are datasets describing the interactions between proteins and molecules. The datasets include a huge number of independent variables (features) and the response variable is typically the binding strength. Thus, one of the most challenging problems is to find the features that have a real influence on the binding strength.

Here we present a feature selection method. The principle of the algorithm is to disturb each single feature by adding pseudo errors and to study the influence on the quality of the model fit.

The main idea is that the change of unimportant features has no effect on the binding strength.

The aim of the project presented in this paper was to investigate the possibility toautomatically sub-classify the superfamily of Short-chain Dehydrogenase/Reductases (SDR).This was done based on an algorithm previously designed to sub-classify the superfamily ofMedium-chain Dehydrogenase/Reductases (MDR). While the SDR-family is interesting andimportant to sub-classify there was also a focus on making the process as automatic aspossible so that future families also can be classified using the same methods.To validate the results generated it was compared to previous sub-classifications done on theSDR-family. The results proved promising and the work conducted here can be seen as a goodinitial part of a more comprehensive full investigation

Copy number variations (CNVs) are not as extensively researched as single nucleotide polymorphisms (SNPs), despite both possessing significant evolutionary potential that could suggest a favored route for adaptation. Through comparative analyses of evolutionary transitions, such as the shift in mating systems from outcrossing to self-fertilization or the comparison of subgenomes following a polyploidization event, general insights into their evolutionary importance can be inferred. For instance, self-fertilizing species tend to accumulate deleterious mutations due to inefficient recombination, unlike outcrossing species. Additionally, a change in ploidy level is anticipated to ease purifying selection on one set of homologous genes while the other set retains its original function. Capsella species, closely related to Arabidopsis thaliana, include four species: Capsella bursa-pastoris (4n, self-fertilizing), Capsella orientalis (2n, selffertilizing), Capsella rubella (2n, self-fertilizing), and Capsella grandiflora (2n, outcrossing). Capsella bursa-pastoris is a model species demonstrating mating system transitions, originating from the hybridization between Capsella orientalis (a self-fertilizer) and the ancestor of Capsella rubella (a self-fertilizer), and Capsella grandiflora (an outcrosser) around 100,000 years ago. The allotetraploid species preserved both genomes, with one having a history of self-fertilization and the other outcrossing. Previous research has determined that the variation in the subgenome of Capsella bursa-pastoris (Cbp) stems from its parental lineages, with some relaxation of purifying selection due to the transition to self-fertilization and the masking of deleterious mutations. Some studies, which consider population history without distinguishing between the two parental subgenomes, have concluded that differences between the populations are attributable to genetic drift.My project aims at testing the hypotheses according to which selfing species accumulates more SVs compared to the outcrossing. By focusing on the gene regions and computing the depth of coverage values of these regions, thresholds are created to classify genes as CNVs. By determining the abundance and presence of the gene CNVs within each subgenome of the different accessions of Cbp, their accumulation is evaluated to draw inferences on the evolutionary trajectory of Cbp.

DNA-binding proteins control how genomes function. The theory of facilitated diffusion(1) explains how DNA-binding proteins can find targets apparently faster than the diffusion limit by using reduced dimensionality(2,3)-combining three-dimensional (3D) diffusion through cytoplasm with 1D sliding along DNA (refs 3-15). However, it does not include a description of macromolecular crowding on DNA as observed in living cells. Here, we show that such a physical constraint to sliding greatly reduces the search speed, in agreement with single-molecule measurements. Interestingly, the generalized theory also reveals significant insights into the design principles of biology. First, it places a hard constraint on the total number of DNA-binding proteins per cell. Remarkably, the number measured for Escherichia coli fits within the optimal range. Secondly, it defines a new role for DNA looping, a ubiquitous topological motif in genomes. DNA looping can speed up the search process by bypassing proteins that block the sliding track close to the target.

A new method for estimation of attributes’ importance for supervised classification, based on the random forest approach, is presented. Essentially, an iterative scheme is applied, with each step consisting of several runs of the random forest program. Each run is performed on a suitably modified data set: values of each attribute found unimportant at earlier steps are randomly permuted between objects. At each step, apparent importance of an attribute is calculated and the attribute is declared unimportant if its importance is not uniformly better than that of the attributes earlier found unimportant. The procedure is repeated until only attributes scoring better than the randomized ones are retained. Statistical significance of the results so obtained is verified. This method has been applied to 12 data sets of biological origin. The method was shown to be more reliable than that based on standard application of a random forest to assess attributes’ importance.

Chemogenomics is a new strategy in in silico drug discovery, where the ultimate goal is to understand molecular recognition for all molecules interacting with all proteins in the proteome. To study such cross interactions, methods that can generalize over proteins that vary greatly in sequence, structure, and function are needed. We present a general quantitative approach to protein−ligand binding affinity prediction that spans the entire structural enzyme-ligand space. The model was trained on a data set composed of all available enzymes cocrystallized with druglike ligands, taken from four publicly available interaction databases, for which a crystal structure is available. Each enzyme was characterized by a set of local descriptors of protein structure that describe the binding site of the cocrystallized ligand. The ligands in the training set were described by traditional QSAR descriptors. To evaluate the model, a comprehensive test set consisting of enzyme structures and ligands was manually curated. The test set contained enzyme-ligand complexes for which no crystal structures were available, and thus the binding modes were unknown. The test set enzymes were therefore characterized by matching their entire structures to the local descriptor library constructed from the training set. Both the training and the test set contained enzyme-ligand complexes from all major enzyme classes, and the enzymes spanned a large range of sequences and folds. The experimental binding affinities (pK_i) ranged from 0.5 to 11.9 (0.7−11.0 in the test set). The induced model predicted the binding affinities of the external test set enzyme-ligand complexes with an r² of 0.53 and an RMSEP of 1.5. This demonstrates that the use of local descriptors makes it possible to create rough predictive models that can generalize over a wide range of protein targets.