This thesis describes improvements to our protein loop structure prediction algorithm and use of this algorithm to inform a computational investigation of anti-HIV antibodies. First, in Section I, we outline improvements to the Protein Local Optimization Program ("Plop") that allow us to reliably restore long loops containing secondary structure elements, and predict nativelike conformations for loops whose surroundings deviate from the native crystal structure context. Shifting to focus exclusively on antibody hypervariable loop prediction, we also benchmark our results in the community-wide Second Antibody Modeling Assessment. Plop can now be reliably deployed as a tool for understanding important biological systems. In Section II, we start from a system of interest - broadly neutralizing antibodies against HIV-1 - with the long-term goal of computationally identifying more potent VRC01-class anti-HIV antibodies. We show proof of concept results for predicting relative binding affinity upon mutation using free energy perturbation (FEP) simulations for the VRC01 antibody binding to glycoprotein gp120. Using the protocols developed in Section I, we provide case studies for using Plop to understand key FEP results and guide future experiments.

Predicting the secondary structure of proteins is important in biochemistry because the 3D structure can be determined from the local folds that are found in secondary structures. Moreover, knowing the tertiary structure of proteins can assist in determining their functions. The objective of this thesis is to compare the performance of Neural Networks (NN) and Support Vector Machines (SVM) in predicting the secondary structure of 62 globular proteins from their primary sequence. For each NN and SVM, we created six binary classifiers to distinguish between the classes’ helices (H) strand (E), and coil (C). For NN we use Resilient Backpropagation training with and without early stopping. We use NN with either no hidden layer or with one hidden layer with 1,2,...,40 hidden neurons. For SVM we use a Gaussian kernel with parameter fixed at = 0.1 and varying cost parameters C in the range [0.1,5]. 10- fold cross-validation is used to obtain overall estimates for the probability of making a correct prediction. Our experiments indicate for NN and SVM that the different binary classifiers have varying accuracies: from 69% correct predictions for coils vs. non-coil up to 80% correct predictions for stand vs. non-strand. It is further demonstrated that NN with no hidden layer or not more than 2 hidden neurons in the hidden layer are sufficient for better predictions. For SVM we show that the estimated accuracies do not depend on the value of the cost parameter. As a major result, we will demonstrate that the accuracy estimates of NN and SVM binary classifiers cannot distinguish. This contradicts a modern belief in bioinformatics that SVM outperforms other predictors.

This thesis encompasses two studies on the application of computational techniques, including deep learning and physics-based methods, in the exploration of protein structure and dynamics. In Chapter 1, I will introduce the background knowledge. Chapter 2 describes the development of a deep learning method for protein loop modeling. We introduce a fast and accurate method for protein loop structure modeling and refinement using deep learning. This method, which is both fast and accurate, integrates a protein language model, a graph neural network, and attention-based modules to predict all-atom protein loop structures from sequences. Its accuracy was validated on benchmark datasets CASP14 and CAMEO, showing performance comparable to or better than the state-of-the-art method, AlphaFold2. The model’s robustness against loop structures outside of the training set was confirmed by testing on datasets after removing high-identity templates and train- ing set homologs. Moreover, it demonstrated significantly lower computational costs compared to existing methods. Application of this method in real-world scenarios included predicting anti- body complementarity-determining regions (CDR) loop structures and refining loop structures in inexact side-chain environments. The method achieved sub-angstrom or near-angstrom accuracy for most CDR loops and notably enhanced the quality of many suboptimal loop predictions in in- exact environments, marking an advancement in protein loop structure prediction and its practical applications. Chapter 3 presents a collaborative study that employs nuclear magnetic resonance (NMR) experiments, molecular dynamics (MD), and hybrid quantum mechanics/molecular mechanics (QM/MM) calculations to investigate protein conformational dynamics across varying temperatures. NMR chemical shifts provide a sensitive probe of protein structure and dynamics. Prediction of shifts, and therefore interpretation of shifts, particularly for the frequently measured amidic 15N sites, remains a tall challenge. We demonstrate that protein ¹⁵N chemical shift prediction from QM/MM predictions can be improved if conformational variation is included via MD sampling, focusing on the antibiotic target, E. coli Dihydrofolate reductase (DHFR). Variations of up to 25 ppm in predicted ¹⁵N chemical shifts are observed over the trajectory. For solution shifts, the average of fluctuations on the low picosecond timescale results in a superior prediction to a single optimal conformation. For low-temperature solid-state measurements, the histogram of predicted shifts for locally minimized snapshots with specific solvent arrangements sampled from the trajectory explains the heterogeneous linewidths; in other words, the conformations and associated solvent are ‘frozen out’ at low temperatures and result in inhomogeneously broadened NMR peaks. We identified conformational degrees of freedom that contribute to chemical shift variation. Backbone torsion angles show high amplitude fluctuations during the trajectory on the low picosecond timescale. For a number of residues, including I60, 𝝍 varies by up to 60o within a conformational basin during the MD simulations, despite the fact that I60 (and other sites studied) are in a secondary structure element and remain well folded during the trajectory. Fluctuations in 𝝍 appear to be compensated by other degrees of freedom in the protein, including 𝝓 of the succeeding residue, resulting in “rocking” of the amide plane with changes in hydrogen bonding interactions. Good agreement for both room-temperature and low-temperature NMR spectra provides strong support for the specific approach to conformational averaging of computed chemical shifts.

Leung Ka-Kit. / Thesis submitted in: November 2004. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2005. / Includes bibliographical references (leaves 67-71). / Abstracts in English and Chinese. / Abstract / Chapter CHAPTER 1. --- INTRODUCTION --- p.1 / Chapter 1.1 --- Background of protein structures --- p.1 / Chapter 1.2 --- Background of protein-protein interaction (PPI) --- p.4 / Chapter 1.2.1 --- Quaternary structure and protein complex --- p.4 / Chapter 1.2.2 --- Previous related work --- p.4 / Chapter 1.2.3 --- The kinetic and thermodynamic formalism --- p.6 / Chapter CHAPTER 2. --- MATERIALS AND METHODS --- p.10 / Chapter 2.1 --- Amino acid composition representative power modeling --- p.10 / Chapter 2.1.1 --- Propensity level modeling --- p.10 / Chapter 2.1.2 --- Polar atoms visualization --- p.17 / Chapter 2.2 --- Rigid structure representative power modeling --- p.17 / Chapter 2.3 --- Electrostatic potential modeling --- p.17 / Chapter 2.3.1 --- Charge residence --- p.17 / Chapter 2.3.2 --- Minimum Ribbon (MR) --- p.19 / Chapter 2.4 --- Examination of interface --- p.23 / Chapter 2.5 --- Identification procedures of a binding site --- p.24 / Chapter 2.6 --- System requirements --- p.24 / Chapter CHAPTER 3. --- RESULTS AND DISCUSSIONS --- p.24 / Chapter 3.1 --- Polar atoms --- p.25 / Chapter 3.2 --- Minimum Ribbon (MR) --- p.27 / Chapter 3.3 --- "Charge complementarity, propensity level and rigid structure orientation" --- p.31 / Chapter 3.4 --- Identification of interacting site --- p.36 / Chapter CHAPTER 4. --- CONCLUSIONS --- p.64 / System requirements --- p.65 / Basic operation --- p.65 / Limitation --- p.66

Fundação Araucária / A Predição da Estrutura das Proteínas, conhecida como PSP (Protein Structure Prediction) pode ser considerada um dos problemas mais desafiadores da Bioinformática atualmente. Quando uma proteína está em seu estado de conformação nativa, a energia livre tende para um valor mínimo. Em geral, a predição da conformação de uma proteína por métodos computacionais é feita pela estimativa de dois valores de energia livre que são provenientes das interações intra e intermoleculares entre os átomos. Alguns estudos recentes indicam que estas interações estão em conflito, justificando o uso de abordagens baseadas em otimização multiobjetivo para a solução do PSP. Neste caso, a otimização destas energias é realizada separadamente, diferente da formulação mono-objetivo que considera a soma das energias. A Evolução Diferencial (ED) é uma técnica baseada em Computação Evolucionária e representa uma alternativa interessante para abordar o PSP. Este trabalho busca desenvolver um otimizador baseado no algoritmo de ED para o problema da Predição da Estrutura de Proteínas multiobjetivo. Este trabalho investiga ainda estratégias baseadas em parâmetros adaptativos para a evolução diferencial. Nicialmente avalia-se uma abordagem simples baseada em ED proposta para a solução do PSP. Uma evolução deste método que incorpora conceitos do algoritmo MOEA/D e adaptação de parâmetros é testada em um conjunto de problemas benchmark de otimização multiobjetivo. Os resultados preliminares obtidos para o PSP (para seis proteínas reais) são promissores e aqueles obtidos para o conjunto benchmark colocam a abordagem proposta como candidata para otimização multiobjetivo. / Protein Structure Prediction (PSP) can be considered one of the most challenging problems in Bioinformatics nowadays. When a protein is in its conformation state, the free energy is minimized. Evaluation of protein conformation is generally performed based on two values of the estimated free energy, i.e., those provided by intra and intermolecular interactions among atoms. Some recent experimental studies show that these interactions are in conflit, justifying the use of multiobjective optimization approaches to solve PSP. In this case, the energy optimization is performed separately, different from the mono-objective optimization which considers the sum of free energy. Differential Evolution (DE) is a technique based on Evolutionary Computation and represents an interesting alternative to solve multiobjective PSP. In this work, an optimizer based on DE is proposed to solve the PSP problem. Due to the great number of parameters, typical for evolutionary algorithms, this work also investigates adaptive parameters strategies. In experiments, a simple approach based on ED is evaluated for PSP. An evolution for this method, which incorporates concepts of the MOEA/D algorithm and parameter adaptation techniques is tested for a set of benchmarks in the multiobjective optimization context. The preliminary results for PSP (for six real proteins) are promising and those obtained for the benchmark set stands the proposed approach as a candidate to the state-of-art for multiobjective optimization.

Fundação Araucária / A Predição da Estrutura das Proteínas, conhecida como PSP (Protein Structure Prediction) pode ser considerada um dos problemas mais desafiadores da Bioinformática atualmente. Quando uma proteína está em seu estado de conformação nativa, a energia livre tende para um valor mínimo. Em geral, a predição da conformação de uma proteína por métodos computacionais é feita pela estimativa de dois valores de energia livre que são provenientes das interações intra e intermoleculares entre os átomos. Alguns estudos recentes indicam que estas interações estão em conflito, justificando o uso de abordagens baseadas em otimização multiobjetivo para a solução do PSP. Neste caso, a otimização destas energias é realizada separadamente, diferente da formulação mono-objetivo que considera a soma das energias. A Evolução Diferencial (ED) é uma técnica baseada em Computação Evolucionária e representa uma alternativa interessante para abordar o PSP. Este trabalho busca desenvolver um otimizador baseado no algoritmo de ED para o problema da Predição da Estrutura de Proteínas multiobjetivo. Este trabalho investiga ainda estratégias baseadas em parâmetros adaptativos para a evolução diferencial. Nicialmente avalia-se uma abordagem simples baseada em ED proposta para a solução do PSP. Uma evolução deste método que incorpora conceitos do algoritmo MOEA/D e adaptação de parâmetros é testada em um conjunto de problemas benchmark de otimização multiobjetivo. Os resultados preliminares obtidos para o PSP (para seis proteínas reais) são promissores e aqueles obtidos para o conjunto benchmark colocam a abordagem proposta como candidata para otimização multiobjetivo. / Protein Structure Prediction (PSP) can be considered one of the most challenging problems in Bioinformatics nowadays. When a protein is in its conformation state, the free energy is minimized. Evaluation of protein conformation is generally performed based on two values of the estimated free energy, i.e., those provided by intra and intermolecular interactions among atoms. Some recent experimental studies show that these interactions are in conflit, justifying the use of multiobjective optimization approaches to solve PSP. In this case, the energy optimization is performed separately, different from the mono-objective optimization which considers the sum of free energy. Differential Evolution (DE) is a technique based on Evolutionary Computation and represents an interesting alternative to solve multiobjective PSP. In this work, an optimizer based on DE is proposed to solve the PSP problem. Due to the great number of parameters, typical for evolutionary algorithms, this work also investigates adaptive parameters strategies. In experiments, a simple approach based on ED is evaluated for PSP. An evolution for this method, which incorporates concepts of the MOEA/D algorithm and parameter adaptation techniques is tested for a set of benchmarks in the multiobjective optimization context. The preliminary results for PSP (for six real proteins) are promising and those obtained for the benchmark set stands the proposed approach as a candidate to the state-of-art for multiobjective optimization.