3.1 Multiple Motif Databases

Two main resources use multiple motifs to characterise protein families: these are PRINTS and Blocks. The scoring method of the former exploits observed residue frequencies, while the latter additionally weights residues not observed in the original alignment.

3.2 Blocks

There are two Blocks databases, one derived automatically from families classified in InterPro, the other from motifs stored in PRINTS (the so-called Blocks-format PRINTS). Blocks based on InterPro families are created automatically using two motif-detection algorithms — motifs detected by both algorithms are considered reliable, otherwise they are discarded. Retained blocks are calibrated against Swiss-Prot to obtain a measure of the chance distribution of matches, and hence of their diagnostic potential. Because the methods used to create these resources differ, so do the chosen motifs — consequently, the same family may be characterised by different blocks within the two databases. As these resources are generated fully automatically, no annotation is provided, but links are provided back to InterPro and PRINTS.

When searching either of the Blocks databases, a rank-ordered list of matches is provided. Good hits usually match more than one block — for multiple-block matches, combined E-values are calculated to reflect the greater significance of the multiple match. High-scoring individual blocks are generally of no relevance.

3.3 PRINTS

PRINTS is a database of manually-crafted protein 'fingerprints'. These are groups of conserved motifs excised from sequence alignments, as shown in Figure 3.9; taken together, they provide diagnostic family signatures. Fingerprints are derived by iterative scanning of a fragmentless Swiss-Prot/TrEMBL composite database — with each database pass, the motifs therefore grow, become more mature and more potently diagnostic of their parent family.

All fingerprints are manually annotated prior to deposition in PRINTS, where they are stored in the form of aligned motifs. Fingerprints inherently offer improved diagnostic reliability over single-motif methods because of the mutual context provided by motif neighbours — results are thus less likely to include false hits, or to miss true ones.

3.4 Challenges for protein family analysis

A strength of fingerprints is their ability to resolve superfamilies into their constituent families and subfamilies. To achieve this, the technique exploits both the similarities between groups of closely related sequences and the differences between them — i.e., between families within a superfamily, or between subfamilies within a family. The choice of motifs is therefore crucial when designing fingerprints at different levels of granularity, as shown in Figure 3.10. A similar challenge occurs with domains families, where incautious choice of motifs would fail to distinguish the domain from of its parent families.

3.5 Database format

Within PRINTS, fingerprints, their match statistics, family annotations and technical details of how they were derived are stored in a single file. For each entry, summaries (and full lists) of both complete and partial fingerprint matches in the given version of Swiss-Prot/TrEMBL are provided.

Query sequences can by searched against the whole of PRINTS, or against individual fingerprints by means of the FingerPRINTScan. This tool returns results on different levels: the first offers the program's 'best guess', based on the highest-scoring full or partial fingerprint match; the next level offers the ten top-scoring hits, allowing users to judge whether significant matches have been missed; a further level provides the full raw search results, rank-ordered according to score. To better understand what these results mean, they may be plotted in the form of a fingerprint 'profiles', which provide at-a-glance views of results and a quick means of judging their likely significance — i.e., if several motifs from a single fingerprint match strongly (above the level of noise) and in the correct order, with appropriate distances between them, this is likely to be a positive diagnosis. E-values provide the measure of significance of multiple motif matches.

3.6 Fingerprint profiles

Fingerprint profiles allow rapid visual diagnosis of results returned by FingerPRINTScan. Within a profile, the x-axis denotes the query sequence, the y-axis the % score for each element of the fingerprint (0-100 per motif); yellow block marks a residue-by-residue match between the query and a particular motif (where the width of the block is the width of the motif and its height the % identity of the match), as shown in Figure 3.11. Even if all motifs are not matched equally well, the pattern of motif formation nevertheless provides a framework for reliable diagnosis.

3.7 Interrogating PRINTS

PRINTS may be interrogated in a variety of ways: e.g., using keywords, Swiss-Prot ID codes, PRINTS ID codes, etc. — the information returned may be modified by changing the Qualifier switch provided. To build more complex queries, simple keyword searches can be combined using the logical operator functions provided by PRINTS's query language — the types of query to be combined may be specified using the Function switch provided.

EMBER: A practical guide to Bioinformatics

EMBER: A practical guide to Bioinformatics