The MPI Bioinformatics Toolkit for protein sequence analysis.

The MPI Bioinformatics Toolkit is an interactive web service which offers access to a great variety of public and in-house bioinformatics tools. They are grouped into different sections that support sequence searches, multiple alignment, secondary and tertiary structure prediction and classification. Several public tools are offered in customized versions that extend their functionality. For example, PSI-BLAST can be run against regularly updated standard databases, customized user databases or selectable sets of genomes. Another tool, Quick2D, integrates the results of various secondary structure, transmembrane and disorder prediction programs into one view. The Toolkit provides a friendly and intuitive user interface with an online help facility. As a key feature, various tools are interconnected so that the results of one tool can be forwarded to other tools. One could run PSI-BLAST, parse out a multiple alignment of selected hits and send the results to a cluster analysis tool. The Toolkit framework and the tools developed in-house will be packaged and freely available under the GNU Lesser General Public Licence (LGPL). The Toolkit can be accessed at http://toolkit.tuebingen.mpg.de.

INTRODUCTION

As this special issue shows, the number of public bioinformatic tools and web servers is growing quickly. However, the wealth of powerful tools and servers is, in our opinion, only utilized by a fraction of biologists who would be able to profit from them. Especially for non-experts it can be very time-consuming to find out which services exist, what they can or cannot do, how to use them and how to feed results from one service to the next in the right format. This has spawned the development of two classes of servers. The first class, exemplified by PredictProtein (1), accepts a single sequence as input, runs a whole set of standard protein analysis tools and returns the bare, concatenated results in a single Email or Web page, requiring users to be familiar with the tools and their output format. The second class offers a collection of web interfaces to local versions of public bioinformatic tools. For instance, PAT (protein analysis toolkit) (2) facilitates the combination of different analysis methods by automating repetitive data processing tasks. However, its user interface and the lack of an integrated help system make PAT, suited primarily for users with biocomputing experience. Two further servers designed as toolboxes for sequence analysis are the Biology Workbench (3), which has not been updated for quite some time, and AnaBench (4), which is more geared toward analysis of DNA data.

The primary aim in developing the MPI Bioinformatics Toolkit was to offer a web service that is as easy to use as possible and that integrates a selected set of most useful methods for the analysis of protein sequences. From our own experience as users of the toolkit, its main advantages are as follows:

In-house tools: Several programs developed in our group are available only through our toolkit, e.g. HHpred (5), HHrep (6), HHsenser (7), REPPER (8), CLANS (9) and Blammer (10) (see Table 1).

Table 1

Overview of tools

Tool	Source references	Description
Search
NucleotideBLAST†	Altschul et al. (11)	Sequence search against nucleotide databases (blastn, tblast, tblastx)
ProteinBLAST†	Altschul et al. (11)	Sequence search against protein databases (blastpgp1, blastx)
PSI-BLAST†	Altschul et al. (12)	Iterated sequence search against protein databases
fastHMMER†	Eddy (13)	Fast profile HMM search tool derived from HMMER
HHpred*	Söding et al. (5)	Sensitive protein homology detection, function and structure prediction by HMM-HMM comparison
HHsenser*	Söding et al. (7)	Sensitive iterative sequence search based on HMM-HMM comparison
PatternSearch*	Unpublished	Search for sequences containing a given pattern
Alignment
ClustalW	Thompson et al. (14)	Multiple alignment program for protein and DNA sequences
MUSCLE	Edgar (16)	Multiple alignment program for protein sequences
ProbCons	Do et al. (15)	Multiple alignment program for protein sequences
MAFFT	Katoh et al. (17)	Multiple alignment program for protein and DNA sequences
Blammer*	Frickey and Lupas (10)	Converts BLAST/PSI-BLAST output to a multiple alignment by realigning gapped regions with Clustal and removing local inconsistencies through comparison to a HMM
HHalign*	Söding (30)	Comparison of two alignments using HMMs
Sequence Analysis
HHrep*	Söding et al. (6)	Sensitive de novo repeat identification in protein sequences by HMM-HMM comparison
PCOILS*	Lupas et al.(31)	Coiled-coil prediction
REPPER*	Gruber et al. (8)	Identification of repeats and their periodicity by Fourier transform and internal sequence comparisons
TPRpred*	Unpublished	Prediction of TPRs (Tetratrico Peptide Repeats) and related repeats (Pentatrico Peptide Repeats and SEL1-like)
Aln2Plot*	Unpublished	Graphical overview of average hydrophobicity and side chain volume in a multiple alignment
Secondary Structure
Quick2D*	Unpublished	Concise overview of secondary structure prediction by PSIPRED (18), JNET (19) and PROFKing (20); of coiled-coils by COILS (31); of transmembrane helices by MEMSAT2 (21) and HMMTOP (22) and of natively disordered regions by DISOPRED2 (23)
Alignment Viewer*	Unpublished	Annotate an alignment with individual PSIPRED (18) and MEMSAT2 (21) predictions
Tertiary Structure
Modeller†	Sali et al. (24)	Comparative protein structure modeling by satisfying of spatial restraints
HHpred*	Söding et al. (5)	Sensitive protein homology detection, function and structure prediction by HMM-HMM comparison
Classification
PHYLIP-NEIGHBOR	Felsenstein (27)	Modules of the phylogenetic analysis package Phylip which allow the construction of distance-based, neighbor-joining trees
CLANS*	Frickey and Lupas (9)	Clustering tool based on all-against-all BLAST comparisons
ANCESCON	Cai et al. (28)	Distance-based phylogenetic inference and reconstruction of ancestral protein sequences
Utilities
Reformat*	Unpublished	Sequence reformatting utility
6FrameTranslation*	Unpublished	Six-frame translation of nucleotide sequences
Extract_gis*	Unpublished	Extraction of gi-numbers from BLAST files
RetrieveSeq*	Unpublished	Sequence retrieval from the nr or nt database using a list of identifiers
gi2Promotor*	Unpublished	Extraction of nucleotide sequences upstream of genes identified by the gi-numbers of their encoded proteins
Backtranslator*	Unpublished	Reverse translation of amino acids into nucleotide sequences

An asterisk after the toolname indicates that the tool was developed in our group.

A dagger indicates a public tool with extended functionality.

Enhanced functionality of public tools: Many tools offer additional functionality compared with the original public server (see tool descriptions below).

User databases: Users may upload customized databases which are then accessible throughout the whole toolkit (upload once, use many times).

Interconnectivity: Most of the tools in the Toolkit are interconnected, allowing job results of one tool to be forwarded as input to others.

Streamlined, uniform user interface: Input forms are kept as simple and self-explanatory as possible with a uniform design and logic for all tools.

Straightforward navigation: Tools are grouped into color-coded sections that are easily accessible via tabs.

Job management: A dedicated jobs sidebar provides information and quick access to all job results of the current session.

Personal work space: Users may register and log in to gain access to a personal work space featuring long-term storage of jobs.

WEB INTERFACE

Currently, 30 bioinformatics tools and utilities can be launched from the MPI Bioinformatics Toolkit (Table 1). All tool sections are accessible from a tabbed menu bar located at the top of the page (Figure 1). Each tab reveals a submenu containing the section-specific tools, an overview page with brief descriptions for each tool and a list of selected links. Located on the left of the screen is a sidebar pane that holds a status and section-coded list of all recent jobs in the current session. One can click on previously submitted jobs to check their status and view their results. Users can also choose their own job names to organize their work. Each tool has a separate input page with a web form, in which the user can input sequence data, upload sequence files, and specify options.

Figure 1

Input and result pages of PSI-BLAST with overlaid windows for genome databases and Jalview alignment viewer (32).

TOOL SECTIONS

The search section contains popular search tools, such as NucleotideBLAST, ProteinBLAST (11), PSI-BLAST (12), and HMMER (13), as well as our in-house developments such as HHpred, HHsenser and PatternSearch. In comparison with the NCBI server, our BLAST tools offer greater flexibility and functionality: searches can be run against uploaded personal databases or selectable sets of genomes (updated weekly from NCBI and ENSEMBL), databases can be switched between PSI-BLAST runs, alignments can be extracted, viewed online or forwarded to other tools, and two graphs show matched regions and E-value distributions. The fastHMMER tool performs HMMER searches of all standard sequence databases in ∼10% of the time by reducing the database with one iteration of PSI-BLAST at a cut-off E-value of 10 000. PatternSearch identifies sequences containing a user-defined Prosite pattern or regular expression. HHpred is a new server for protein structure and function prediction (5). It takes a query sequence as input and searches user-selected databases for homologs with a new and very sensitive method based on pairwise comparison of hidden Markov models (HMMs). Available databases, among others, are InterPro, CDD and an aligment database we build from Protein Data Bank (PDB) sequences and which can be used for 3D structure prediction. HHsenser is a transitive search method based on HMM-HMM comparison (7). This method utilizes a sequence as input and builds an alignment with as many near or remote homologs as possible, often covering the whole protein superfamily.

The alignment section includes the well-known, popular multiple alignment program ClustalW (14), together with the more recently developed multiple alignment methods ProbCons (15), MUSCLE (16) and MAFFT (17). Also in this section is Blammer (10), which converts BLAST or PSI-BLAST output to a multiple alignment by realigning gapped regions using ClustalW and removing local inconsistencies through comparison with an HMM. HHalign aligns two alignments with each other by pairwise comparison of HMMs and displays similarities in a profile–profile dotplot.

In the sequence analysis section, we have grouped tools for repeat identification and analysis of periodic regions in proteins. HHrep is a server for de novo repeat detection that is very sensitive in finding proteins with strongly diverged repeats, such as TIM barrels and β-propellers (6). REPPER (8) analyzes regions with short gapless repeats in protein sequences. It finds periodicities by Fourier transform and internal sequence similarity. The output is complemented by coiled-coil prediction and secondary structure prediction using PSIPRED (18). Aln2Plot shows a graphical overview of average hydrophobicity and side chain volume in a multiple alignment.

In the secondary structure section, Quick2D integrates the results of various secondary structure prediction programs, such as PSIPRED (18), JNET (19) and PROFKing (20), the transmembrane prediction of MEMSAT2 (21) and HMMTOP (22) and the disorder prediction of DISOPRED (23) into a single colored view. The AlignmentViewer clusters sequences by a sequence idenity criterion, annotates groups of sequences using PSIPRED and MEMSAT2 predictions of a multiple alignment and graphically displays the results in an interactive Java applet.

The tertiary structure section contains Modeller (24) and HHpred (5). Modeller is a very popular program for comparative modeling. It generates a 3D structural model from a sequence alignment of a protein sequence with one or more structural templates. In contrast to the standalone version of Modeller, the input format does not need to be PIR but can also be FASTA or most other standard multiple alignment formats. Modeller is tightly integrated with HHpred, allowing selected hits of HHpred results to be used as templates for subsequent comparative modeling. On the results page, models can be evaluated by using a browser-embedded 3D-viewer and charts with output from several model quality assessment programs are provided. This allows fast interactive refinement cycles of the underlying multiple sequence alignment. The page also provides a link to the iMolTalk server, which offers several additional tools for the detailed analysis of structures and models (25,26).

In the classification section, we offer modules of the widely used phylogenetic analysis suite PHYLIP (27), the ANCESCON package (28) for distance bases phylogenetic analysis and CLANS (9). CLANS clusters user-provided sequences based on BLAST pairwise similarities (29). The results can be analysed with a CLANS Java applet or can br exported to CLANS format.

Finally, in the utilities section there is a collection of tools which help to perform simple tasks that the user will often be confronted with. It includes a sequence reformatting utility, a six-frame translation tool for nucleotide sequences, Extract_gis for the extraction of gi-numbers from BLAST files, the RetrieveSeq tool for identifier-based sequence retrieval from the non-redundant protein or nucleotide databases at NCBI, gi2Promotor for the extraction of nucleotide sequences upstream of genes identified by the gi-numbers of their encoded proteins and a backtranslation tool.

FUTURE PLANS

Our own research on protein evolution now heavily depends on the toolkit server. We will therefore continue to integrate new tools as they become available and improve the usability of the toolkit. For instance, a project manager will be added that will further facilitate the organization and long-term storage of job results. On the technical side, we are currently in the process of porting the Toolkit to a new Rails-based web framework that permits shorter development cycles and more flexible tool interactions. The new architecture is fully object oriented and renders the Toolkit easily installable. We will package the Toolkit framework together with our in-house tools and distribute it freely under the GNU LGPL.