Protein Structure and Function Prediction Methods

MetaServer
The structure prediction Meta Server offers a gateway to many high quality fold recognition servers and provides an infrastructure and main interface to several highly reliable consensus methods. The Meta Server represents the first step in our fold prediction strategy.

LiveBench
The Live Bench Project is a continuous benchmarking program. Every week sequences of newly released PDB proteins are being submitted to participating fold recognition servers. The results are collected and continuous evaluated using automated model assessment programs. A summary of the results is produced after several months of data collection. The servers must delay the updating of their structural template labraries by one week to participate.

GRDB
The Gene Relational Database facilitates the comparison of protein families using BASIC. BASIC is a novel sensitive approach for recognition of distant similarity between proteins based on consensus alignments of meta profiles. Specifically, BASIC compares sequence profiles combined with predicted secondary structure by utilizing several scoring systems and alignment algorithms. In our benchmarking tests, BASIC outperforms many individual servers, including fold recognition servers, and it can compete with meta predictors that base their strength on the structural comparison of models. In addition, BASIC, which enables detection of very distant relationships even if the tertiary structure for the reference protein is not known, has high-throughput capability. GRDB operates on a grid of computers placed in different institutions. TOM is used for task management.

AutoMotif
The AutoMotif Server (AMS) predicts functional motifs in proteins based only on sequence information. A list of possible functional sites for a given query protein is constructed using its sequence and the database of proteins annotated for a certain type of biochemical process by Swiss-Prot database. The whole query protein sequence is dissected into overlapping short segments. All segments are then projected into one abstract space of sequence fragments by different representations (10 different embeddings). Those representations are compared with the database of representations of known functional motifs using the support vector machine SVM approach. The efficiency of the classification for each type of active site and the prediction power of the method is estimated using the leave-one-out tests and presented here. Registered users can access all sites annotated by Swiss-Prot database (version 4.2), add new proteins with annotated segments (positive instances) or change attributes of already included proteins. All data, biological information, theoretical classification models and automatic functional predictor are updated after each major upgrade of the Swiss-Prot DB.

Small Molecules Tools

Cancer Drug Server
This service facilitates the analysis of correlation between the activity of many chemical substances tested in (NCI60) anti-cancer trials and the expression of genes deduced from cDNA microarray results obtained for 60 cancer cell lines. Because both types of experiments were conducted on the same group of cells the results can be directly compared with each other. On one hand it may seem unlikely to find a gene significantly affecting the action of a drug from a list of almost 10 000 investigated human genes just by collecting 60 measurements. On the other hand if each measurement could provide just 2 possible values the total number of possible results would be 60^2 = 1024^6 a number with 19 figures, much larger than the number of nucleotides in gene sequences stored in current databases. Thus, 60 experiments are sufficient to create a substance-specific of gene-specific activity profile. The results can point to interesting genes or groups of genes interacting in some way with the compound of interest.

LigandInfo
Ligand.Info is a system of Small-Molceule Database and Java-based tool for virtual high-throughput screening of new potential drugs. The Ligand.Info Meta-Database contains various publicly available sets of small molecules such as Harvard's ChemBank, Hetero Atoms from Protein Data Bank, KEGG Ligand Database, The Open NCI Database, and others. The total size of the database is 1 million entries. The Ligand.Info server is based on the idea that small molecules with similar structure have similar biological properties. The developed system enables a fast and sensitive search for similar compounds using structural indices. The tool can interactively cluster sets of molecules on the user side and automatically download similar molecules from the Meta-Database. All downloaded molecules are automatically displayed using the structure viewer.