Novel Computational Protein Design Algorithms with Applications to Cystic Fibrosis and HIV

Proteins are essential components of cells and are crucial for catalyzing reactions, signaling, recognition, motility, recycling, and structural stability. This diversity of function suggests that nature is only scratching the surface of protein functional space. Protein function is determined by structure, which in turn is determined predominantly by amino acid sequence. Protein design aims to explore protein sequence and conformational space to design novel proteins with new or improved function. The vast number of possible protein sequences makes exploring the space a challenging problem.

Computational structure-based protein design (CSPD) allows for the rational design of proteins. Because of the large search space, CSPD methods must balance search accuracy and modeling simplifications. We have developed algorithms that allow for the accurate and efficient search of protein conformational space. Specifically, we focus on algorithms that maintain provability, account for protein flexibility, and use ensemble-based rankings. We present several novel algorithms for incorporating improved flexibility into CSPD with continuous rotamers. We applied these algorithms to two biomedically important design problems. We designed peptide inhibitors of the cystic fibrosis agonist CAL that were able to restore function of the vital cystic fibrosis protein CFTR. We also designed improved HIV antibodies and nanobodies to combat HIV infections.

Computational modeling of mold filling and related free-surface flows in shape casting: an overview of the challenges involved

Accurate representation of the coupled effects between turbulent fluid flow with a free surface, heat transfer, solidification, and mold deformation has been shown to be necessary for the realistic prediction of several defects in castings and also for determining the final crystalline structure. A core component of the computational modeling of casting processes involves mold filling, which is the most computationally intensive aspect of casting simulation at the continuum level. Considering the complex geometries involved in shape casting, the evolution of the free surface, gas entrapment, and the entrainment of oxide layers into the casting make this a very challenging task in every respect. Despite well over 30 years of effort in developing algorithms, this is by no means a closed subject. In this article, we will review the full range of computational methods used, from unstructured finite-element (FE) and finite-volume (FV) methods through fully structured and block-structured approaches utilizing the cut-cell family of techniques to capture the geometric complexity inherent in shape casting. This discussion will include the challenges of generating rapid solutions on high-performance parallel cluster technology and how mold filling links in with the full spectrum of physics involved in shape casting. Finally, some indications as to novel techniques emerging now that can address genuinely arbitrarily complex geometries are briefly outlined and their advantages and disadvantages are discussed.

A Virtual Inspection Technique for Assessing the Dimensional Accuracy of Forged Compressor Blades Using FE Modeling and CMM Inspection

This paper presents research for developing a virtual inspection system that evaluates the dimensional tolerance of forged aerofoil blades formed using the finite element (FE) method. Conventional algorithms adopted by modern coordinate measurement processes have been incorporated with the latest free-form surface evaluation techniques to provide a robust framework for the dimensional inspection of FE aerofoil models. The accuracy of the approach had been verified with a strong correlation obtained between the virtual inspection data and coordinate measurement data from corresponding aerofoil components.

Analysis, modeling, design and optimization of future communications systems: from theory to practice

Esta tese descreve uma framework de trabalho assente no paradigma multi-camada para analisar, modelar, projectar e optimizar sistemas de comunicação. Nela se explora uma nova perspectiva acerca da camada física que nasce das relações entre a teoria de informação, estimação, métodos probabilísticos, teoria da comunicação e codificação. Esta framework conduz a métodos de projecto para a próxima geração de sistemas de comunicação de alto débito. Além disso, a tese explora várias técnicas de camada de acesso com base na relação entre atraso e débito para o projeto de redes sem fio tolerantes a atrasos. Alguns resultados fundamentais sobre a interação entre a teoria da informação e teoria da estimação conduzem a propostas de um paradigma alternativo para a análise, projecto e optimização de sistemas de comunicação. Com base em estudos sobre a relação entre a informação recíproca e MMSE, a abordagem descrita na tese permite ultrapassar, de forma inovadora, as dificuldades inerentes à optimização das taxas de transmissão de informação confiáveis em sistemas de comunicação, e permite a exploração da atribuição óptima de potência e estruturas óptimas de pre-codificação para diferentes modelos de canal: com fios, sem fios e ópticos. A tese aborda também o problema do atraso, numa tentativa de responder a questões levantadas pela enorme procura de débitos elevados em sistemas de comunicação. Isso é feito através da proposta de novos modelos para sistemas com codificação de rede (network coding) em camadas acima da sua camada física. Em particular, aborda-se a utilização de sistemas de codificação em rede para canais que variam no tempo e são sensíveis a atrasos. Isso foi demonstrado através da proposta de um novo modelo e esquema adaptativo, cujos algoritmos foram aplicados a sistemas sem fios com desvanecimento (fading) complexo, de que são exemplos os sistemas de comunicação via satélite. A tese aborda ainda o uso de sistemas de codificação de rede em cenários de transferência (handover) exigentes. Isso é feito através da proposta de novos modelos de transmissão WiFi IEEE 801.11 MAC, que são comparados com codificação de rede, e que se demonstram possibilitar transferência sem descontinuidades. Pode assim dizer-se que esta tese, através de trabalho de análise e de propostas suportadas por simulações, defende que na concepção de sistemas de comunicação se devem considerar estratégias de transmissão e codificação que sejam não só próximas da capacidade dos canais, mas também tolerantes a atrasos, e que tais estratégias têm de ser concebidas tendo em vista características do canal e a camada física.

Comparison of on-line learning algorithms for RBF models in greenhouse environmental control problems

The problem with the adequacy of radial basis function neural networks to model the inside air temperature as a function of the outside air temperature and solar radiation, and the inside relative humidity in an hydroponic greenhouse is addressed.

Fuzzy model identification by evolutionary, gradient based and memtic algorithms

One of the crucial problems of fuzzy rule modeling is how to find an optimal or at least a quasi-optimal rule base fro a certain system. In most applications there is no human expert available, or, the result of a human expert's decision is too much subjective and is not reproducible, thus some automatic method to determine the fuzzy rule base must be deployed.

Modeling the suitable habitat of a structural red gorgonian species in the Mediterranean/east Atlantic regions predicts a wider distribution than previously known

Dissertação de mestrado, Biologia Marinha, Faculdade de Ciências e Tecnologia, Univerdade do Algarve, 2015

Machine learning methods for quantitative structure-property relationship modeling

Tese de doutoramento, Informática (Bioinformática), Universidade de Lisboa, Faculdade de Ciências, 2014

Gene Network Inference using Machine Learning and Graph Algorithms on Big Biomedical Data

Thesis (Master's)--University of Washington, 2016-03

Modeling sequencing errors by combining Hidden Markov models.

Among the largest resources for biological sequence data is the large amount of expressed sequence tags (ESTs) available in public and proprietary databases. ESTs provide information on transcripts but for technical reasons they often contain sequencing errors. Therefore, when analyzing EST sequences computationally, such errors must be taken into account. Earlier attempts to model error prone coding regions have shown good performance in detecting and predicting these while correcting sequencing errors using codon usage frequencies. In the research presented here, we improve the detection of translation start and stop sites by integrating a more complex mRNA model with codon usage bias based error correction into one hidden Markov model (HMM), thus generalizing this error correction approach to more complex HMMs. We show that our method maintains the performance in detecting coding sequences.

Modeling Metal Protein Complexes from Experimental Extended X-ray Absorption Fine Structure using Computational Intelligence

Experimental Extended X-ray Absorption Fine Structure (EXAFS) spectra carry information about the chemical structure of metal protein complexes. However, pre- dicting the structure of such complexes from EXAFS spectra is not a simple task. Currently methods such as Monte Carlo optimization or simulated annealing are used in structure refinement of EXAFS. These methods have proven somewhat successful in structure refinement but have not been successful in finding the global minima. Multiple population based algorithms, including a genetic algorithm, a restarting ge- netic algorithm, differential evolution, and particle swarm optimization, are studied for their effectiveness in structure refinement of EXAFS. The oxygen-evolving com- plex in S1 is used as a benchmark for comparing the algorithms. These algorithms were successful in finding new atomic structures that produced improved calculated EXAFS spectra over atomic structures previously found.

Timing verification in transaction modeling

Les systèmes Matériels/Logiciels deviennent indispensables dans tous les aspects de la vie quotidienne. La présence croissante de ces systèmes dans les différents produits et services incite à trouver des méthodes pour les développer efficacement. Mais une conception efficace de ces systèmes est limitée par plusieurs facteurs, certains d'entre eux sont: la complexité croissante des applications, une augmentation de la densité d'intégration, la nature hétérogène des produits et services, la diminution de temps d’accès au marché. Une modélisation transactionnelle (TLM) est considérée comme un paradigme prometteur permettant de gérer la complexité de conception et fournissant des moyens d’exploration et de validation d'alternatives de conception à des niveaux d’abstraction élevés. Cette recherche propose une méthodologie d’expression de temps dans TLM basée sur une analyse de contraintes temporelles. Nous proposons d'utiliser une combinaison de deux paradigmes de développement pour accélérer la conception: le TLM d'une part et une méthodologie d’expression de temps entre différentes transactions d’autre part. Cette synergie nous permet de combiner dans un seul environnement des méthodes de simulation performantes et des méthodes analytiques formelles. Nous avons proposé un nouvel algorithme de vérification temporelle basé sur la procédure de linéarisation des contraintes de type min/max et une technique d'optimisation afin d'améliorer l'efficacité de l'algorithme. Nous avons complété la description mathématique de tous les types de contraintes présentées dans la littérature. Nous avons développé des méthodes d'exploration et raffinement de système de communication qui nous a permis d'utiliser les algorithmes de vérification temporelle à différents niveaux TLM. Comme il existe plusieurs définitions du TLM, dans le cadre de notre recherche, nous avons défini une méthodologie de spécification et simulation pour des systèmes Matériel/Logiciel basée sur le paradigme de TLM. Dans cette méthodologie plusieurs concepts de modélisation peuvent être considérés séparément. Basée sur l'utilisation des technologies modernes de génie logiciel telles que XML, XSLT, XSD, la programmation orientée objet et plusieurs autres fournies par l’environnement .Net, la méthodologie proposée présente une approche qui rend possible une réutilisation des modèles intermédiaires afin de faire face à la contrainte de temps d’accès au marché. Elle fournit une approche générale dans la modélisation du système qui sépare les différents aspects de conception tels que des modèles de calculs utilisés pour décrire le système à des niveaux d’abstraction multiples. En conséquence, dans le modèle du système nous pouvons clairement identifier la fonctionnalité du système sans les détails reliés aux plateformes de développement et ceci mènera à améliorer la "portabilité" du modèle d'application.

Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

Inverse problems in dynamic cognitive modeling

Inverse problems for dynamical system models of cognitive processes comprise the determination of synaptic weight matrices or kernel functions for neural networks or neural/dynamic field models, respectively. We introduce dynamic cognitive modeling as a three tier top-down approach where cognitive processes are first described as algorithms that operate on complex symbolic data structures. Second, symbolic expressions and operations are represented by states and transformations in abstract vector spaces. Third, prescribed trajectories through representation space are implemented in neurodynamical systems. We discuss the Amari equation for a neural/dynamic field theory as a special case and show that the kernel construction problem is particularly ill-posed. We suggest a Tikhonov-Hebbian learning method as regularization technique and demonstrate its validity and robustness for basic examples of cognitive computations.