Adaptive-BLAST: A user-defined platform for the study of proteins

Profile-based protein-sequence analysis algorithms comprise some of the most powerful and user-friendly methods for exploring protein sequences to determine their structure, function, and/or evolution (1-4). PSI-BLAST (5, 6) and rps-BLAST (7) are two of the most popular profile-based algorithms (~1,120 references to date), and have exceptional utility in the identification of homology between proteins, particularly for biological scientists who do not specialize in computational approaches. However, when the performance of these algorithms is compared to other methods [e.g. support-vector machine learning (SVM) (8), hidden-Markov models (HMMs) (9)], they often underperform in identifying the aforementioned protein properties (3, 9-11). We have previously demonstrated that the utility of BLAST algorithms can be significantly improved by: (i) adaptations to the profile libraries employed, (ii) adjustments to output formats, and (iii) alterations to BLAST algorithm itself (4, 6, 12-14). We present here Adaptive-BLAST (Ada-BLAST), which provides a simple user-defined platform for measuring and analyzing primary amino acid sequences. Within this platform, we developed a series of local BLAST applications (apps) that take advantage of the speed and sensitivity afforded by BLAST, while allowing for maximal user-definitions and flexible visualization. We tested the efficacy of these apps in control experiments, studying fold-recognition, in which we obtained >90% accuracy in highly divergent sequences (>25% identity). In addition, these same apps were proficient in classifying transmembrane proteins, identifying structural/functional determinants of ion-channels/receptors, and informing structural modeling algorithms. Indeed, these Ada-BLAST informed-structural models were useful in guiding our experimental research on the N-terminus of Transient Receptor Potential ion-channels (TRPs). Taken together, we propose that Ada-BLAST provides a powerful computational tool that is accessible to bench-scientists and computational biologists alike. The codes for Ada-BLAST are publicly available at: http://empathy.rcc.psu.edu/.