Accuracy Estimation and Parameter Advising for Protein Multiple Sequence Alignment

Abstract

We develop a novel and general approach to estimating the accuracy of multiple sequence alignments without knowledge of a reference alignment, and use our approach to address a new task that we call parameter advising: the problem of choosing values for alignment scoring function parameters from a given set of choices to maximize the accuracy of a computed alignment.For protein alignments, we consider twelve independent features that contribute to a quality alignment. An accuracy estimator is learned that is a polynomial function of these features; its coefficients are determined by minimizing its error with respect to true accuracy using mathematical optimization. Compared to prior approaches for estimating accuracy, our new approach (a) introduces novel feature functions that measure nonlocal properties of an alignment yet are fast to evaluate, (b) considers more general classes of estimators beyond linear combinations of features, and (c) develops new regression formulations for learning an estimator from examples; in addition, for parameter advising, we (d) determine the optimal parameter set of a given cardinality, which specifies the best parameter values from which to choose.Our estimator, which we call Facet (for ''feature-based accuracy estimator''), yields a parameter advisor that on the hardest benchmarks provides more than a 27% improvement in accuracy over the best default parameter choice, and for parameter advising significantly outperforms the best prior approaches to assessing alignment quality.

Ensemble multiple sequence alignment via advising

“…Wheeler and Kececioglu [34] first introduced the notion of parameter advisors; Kececioglu and DeBlasio [14] investigated the construction of alignment accuracy estimators, resulting in the Facet estimator [4]; DeBlasio and Kececioglu [6] investigated how to best form the set of parameter choices for an advisor, called an advisor set, developing an efficient approximation algorithm for finding a near-optimal advisor set for a given estimator. This prior work applied parameter advising to boosting the accuracy of the Opal aligner [35].…”

“…The AQUA tool chooses between an alignment computed by MUSCLE [9] or MAFFT [13] based on their NorMD [31] score; our prior work [14] shows that for choosing the more accurate alignment, the NorMD score used by AQUA is much weaker than the Facet estimator used here for aligner advising. M-Coffee uses a standard progressive alignment heuristic to compute an alignment under positiondependent substitution scores whose values are determined by alignments from different aligners.…”

“…Each point in a scatterplot corresponds to an alignment whose true accuracy is on the horizontal axis, and whose value under a given estimator is on the vertical axis. Both scatterplots show the same set of 3,000 alignments (randomly sampled from the more than 209,000 alignments generated by the experiments in Section 4) under the accuracy estimators Facet [14] and TCS [3].…”

“…Kececioglu and DeBlasio [7,14] present an efficient approach for learning an accuracy estimator that is a linear combination of real-valued alignment feature functions, based on solving a large-scale linear programming problem. This approach resulted in the Facet estimator [4], which is currently the most accurate estimator for parameter advising [14,6].…”

“…This approach resulted in the Facet estimator [4], which is currently the most accurate estimator for parameter advising [14,6].…”