The Sorting Intolerant from Tolerant (SIFT) algorithm predicts the effect of coding variants on protein function. It was first introduced in 2001, with a corresponding website that provides users with predictions on their variants. Since its release, SIFT has become one of the standard tools for characterizing missense variation. We have updated SIFT’s genome-wide prediction tool since our last publication in 2009, and added new features to the insertion/deletion (indel) tool. We also show accuracy metrics on independent data sets. The original developers have hosted the SIFT web server at FHCRC, JCVI and the web server is currently located at BII. The URL is http://sift-dna.org (24 May 2012, date last accessed).

Each human has approximately 50 to 280 frameshifting indels, yet their implications are unknown. We created SIFT Indel, a prediction method for frameshifting indels that has 84% accuracy. The percentage of human frameshifting indels predicted to be gene-damaging is negatively correlated with allele frequency. We also show that although the first frameshifting indel in a gene causes loss of function, there is a tendency for the second frameshifting indel to compensate and restore protein function. SIFT Indel is available at http://sift-dna.org/www/SIFT_indels2.html

Indels in the coding regions of a gene can either cause frameshifts or amino acid insertions/deletions. Frameshifting indels are indels that have a length that is not divisible by 3 and subsequently cause frameshifts. Indels that have a length divisible by 3 cause amino acid insertions/deletions or block substitutions; we call these 3n indels. The new amino acid changes resulting from 3n indels could potentially affect protein function. Therefore, we construct a SIFT Indel prediction algorithm for 3n indels which achieves 82% accuracy, 81% sensitivity, 82% specificity, 82% precision, 0.63 MCC, and 0.87 AUC by 10-fold cross-validation. We have previously published a prediction algorithm for frameshifting indels. The rules for the prediction of 3n indels are different from the rules for the prediction of frameshifting indels and reflect the biological differences of these two different types of variations. SIFT Indel was applied to human 3n indels from the 1000 Genomes Project and the Exome Sequencing Project. We found that common variants are less likely to be deleterious than rare variants. The SIFT indel prediction algorithm for 3n indels is available at http://sift-dna.org/

No abstract

This paper presents a parameter-free integer-programming based algorithm for the global resolution of a linear program with linear complementarity constraints (LPCC). The cornerstone of the algorithm is a minimax integer program formulation that characterizes and provides certificates for the three outcomes-infeasibility, unboundedness, or solvability-of an LPCC. An extreme point/ray generation scheme in the spirit of Benders decomposition is developed, from which valid inequalities in the form of satisfiability constraints are obtained. The feasibility problem of these inequalities and the carefully guided linear programming relaxations of the LPCC are the workhorse of the algorithm, which also employs a specialized procedure for the sparsification of the satifiability cuts. We establish the finite termination of the algorithm and report computational results using the algorithm for solving randomly generated LPCCs of reasonable sizes. The results establish that the algorithm can handle infeasible, unbounded, and solvable LPCCs effectively.

Support vector machines and related classification models require the solution of convex optimization problems that have one or more regularization hyper-parameters. Typically, the hyper-parameters are selected to minimize the cross-validated estimates of the out-of-sample classification error of the model. This cross-validation optimization problem can be formulated as a bilevel program in which the outer-level objective minimizes the average number of misclassified points across the cross-validation folds, subject to inner-level constraints such that the classification functions for each fold are (exactly or nearly) optimal for the selected hyper-parameters. Feature selection is included in the bilevel program in the form of bound constraints in the weights. The resulting bilevel problem is converted to a mathematical program with linear equilibrium constraints, which is solved using state-of-the-art optimization methods. This approach is significantly more versatile than commonly used grid search procedures, enabling, in particular, the use of models with many hyper-parameters. Numerical results demonstrate the practicality of this approach for model selection in machine learning.

No abstract

No abstract