MOTIVATION: Beta turns are the most common type of secondary structure in proteins after alpha helices and beta sheets and play many key structural and functional roles. Turn backbone (BB) geometry has been classified at multiple levels of precision, but the current picture of side chain (SC) structure and interaction in turns is incomplete, because the distribution of SC conformations associated with each sequence motif has commonly been represented only by a static image of a single, typical structure for each turn BB geometry, and only motifs which specify a single amino acid (e.g. aspartic acid at turn position 1) have been systematically investigated. Furthermore, no general evaluation has been made of the SC interactions between turns and their BB neighborhoods. Finally, the visualization and comparison of the wide range of turn conformations has been hampered by the almost exclusive characterization of turn structure in BB dihedral-angle (Ramachandran) space. RESULTS: This work introduces MapTurns, a web server for motif maps, which employ a turn-local Euclidean-space coordinate system and a global turn alignment to comprehensively map the distributions of BB and SC structure and H-bonding associated with sequence motifs in beta turns and their local BB contexts. Maps characterize many new SC motifs, provide detailed rationalizations of sequence preferences, and support mutational analysis and the general study of SC interactions, and they should prove useful in applications such as protein design. AVAILABILITY AND IMPLEMENTATION: MapTurns is available at www.betaturn.com. Sample code is available at: https://github.com/nenewell/MapTurns/tree/main.

BACKGROUND: Interactions that involve one or more amino acid side chains near the ends of protein helices stabilize helix termini and shape the geometry of the adjacent loops, making a substantial contribution to overall protein structure. Previous work has identified key helix-terminal motifs, such as Asx/ST N-caps, the capping box, and hydrophobic and electrostatic interactions, but important questions remain, including: 1) What loop backbone geometries are favoured by each motif? 2) To what extent are multi-amino acid motifs likely to represent genuine cooperative interactions? 3) Can new motifs be identified in a large, recent dataset using the latest bioinformatics tools? RESULTS: Three analytical tools are applied here to answer these questions. First, helix-terminal structures are partitioned by loop backbone geometry using a new 3D clustering algorithm. Next, Cascade Detection, a motif detection algorithm recently published by the author, is applied to each cluster to determine which sequence motifs are overrepresented in each geometry. Finally, the results for each motif are presented in a CapMap, a 3D conformational heatmap that displays the distribution of the motif's overrepresentation across loop geometries, enabling the rapid isolation and characterization of the associated side chain interaction. This work identifies a library of geometry-specific side chain interactions that provides a new, detailed picture of loop structure near the helix terminus. Highlights include determinations of the favoured loop geometries for the Asx/ST N-cap motifs, capping boxes, "big" boxes, and other hydrophobic, electrostatic, H-bond, and pi stacking interactions, many of which have not been described before. CONCLUSIONS: This work demonstrates that the combination of structural clustering and motif detection in the sequence space can efficiently identify side chain motifs and map them to the loop geometries which they support. Protein designers should find this study useful, because it identifies side chain interactions which are good candidates for inclusion in synthetic helix-terminal loops with specific desired geometries, since they are used in nature to support these geometries. The techniques described here can also be applied to map side chain interactions associated with other structural components of proteins such as beta and gamma turns.

MOTIVATION: The extraction of the set of features most relevant to function from classified biological sequence sets is still a challenging problem. A central issue is the determination of expected counts for higher order features so that artifact features may be screened. RESULTS: Cascade detection (CD), a new algorithm for the extraction of localized features from sequence sets, is introduced. CD is a natural extension of the proportional modeling techniques used in contingency table analysis into the domain of feature detection. The algorithm is successfully tested on synthetic data and then applied to feature detection problems from two different domains to demonstrate its broad utility. An analysis of HIV-1 protease specificity reveals patterns of strong first-order features that group hydrophobic residues by side chain geometry and exhibit substantial symmetry about the cleavage site. Higher order results suggest that favorable cooperativity is weak by comparison and broadly distributed, but indicate possible synergies between negative charge and hydrophobicity in the substrate. Structure-function results for the Schellman loop, a helix-capping motif in proteins, contain strong first-order features and also show statistically significant cooperativities that provide new insights into the design of the motif. These include a new 'hydrophobic staple' and multiple amphipathic and electrostatic pair features. CD should prove useful not only for sequence analysis, but also for the detection of multifactor synergies in cross-classified data from clinical studies or other sources. AVAILABILITY: Windows XP/7 application and data files available at: https://sites.google.com/site/cascadedetect/home. CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary information is available at Bioinformatics online.