Prediction of Protein Retention Times in Anion-Exchange Chromatography Systems Using Support Vector Regression

Abstract

Quantitative Structure-Retention Relationship (QSRR) models are developed for the prediction of protein retention times in anion-exchange chromatography systems. Topological, subdivided surface area, and TAE (Transferable Atom Equivalent) electron-density-based descriptors are computed directly for a set of proteins using molecular connectivity patterns and crystal structure geometries. A novel algorithm based on Support Vector Machine (SVM) regression has been employed to obtain predictive QSRR models using a two-step computational strategy. In the first step, a sparse linear SVM was utilized as a feature selection procedure to remove irrelevant or redundant information. Subsequently, the selected features were used to produce an ensemble of nonlinear SVM regression models that were combined using bootstrap aggregation (bagging) techniques, where various combinations of training and validation data sets were selected from the pool of available data. A visualization scheme (star plots) was used to display the relative importance of each selected descriptor in the final set of "bagged" models. Once these predictive models have been validated, they can be used as an automated prediction tool for virtual high-throughput screening (VHTS).

An Introduction to the Quantum Theory of Atoms in Molecules

“…Atomic properties have also been used empirically to predict several experimental properties including for example, the pK a of weak acids from the atomic energy of the acidic hydrogen [96], a wide array of biological and physicochemical properties of the amino acids, including the genetic code itself, and the effects of mutation on protein stability [60], protein retention times [97], HPLC column capacity factors of high-energy materials [98], NMR spin-spin coupling constants from the electron delocalization indices [99,100], simultaneous consistent prediction of five bulk properties of liquid HF in MD simulation [101], classification of atom types in proteins with future potential applications in force-field design [60,[102][103][104], reconstructing large molecules from transferable fragments or atoms in molecules [60,[105][106][107][108][109][110][111][112][113][114][115][116][117][118][119] (see also Chapters 11 and 12), atomic partitioning of the molecular electrostatic potential [120][121][122], prediction of hydrogenbond donor capacity [123] and basicity [124], and to provide an atomic basis for curvature-induced polarization in carbon nanotubes and nanoshells [125].…”

High Throughput Screening for the Design and Optimization of Chromatographic Processes: Automated Optimization of Chromatographic Phase Systems

“…The development of mechanistic models requires good process understanding, initial experiments for parameter estimation, and independent experiments for model validation. The latter is also true for black box models (for example [1,2]). The determination of mechanistic model parameters, such as effective mass transfer coefficients and isotherm coefficients, is generally very complex and requires large amounts of material and time, especially when the interactions of realistic multi-component mixtures are considered without significant model simplifications.…”

“…(1)(2)(3)(4) show that asymmetric Gaussian distributions are uniquely characterized by the following four parameters: peak area, Ar, location of peak maximum, l, average peak width, r, and peak asymmetry, a. These parameters are estimated from measurement data so as to minimize the least-squares distance between chromatograms and the sum of several asymmetric Gaussian peaks:…”

Diagnosis of breast tumors with ultrasonic texture analysis using support vector machines

“…The SVMs have been shown to be an efficient method for many real-world problems because of its high generalization performance without the need to add a priori knowledge. Thus, SVMs have much attention as a successful tool for classification [12,13], image recognition [14,15] and bioinformatics [16,17]. The SVM model can map the input vectors into a high-dimensional feature space through some non-linear mapping, chosen a priori.…”