No description available.

<p>This thesis is concerned with the prediction of protein mutations using artificial neural networks. From the biological perspective it is of interest to investigate weather it is possible to find rules of mutation between evolutionary adjacent (or closely related) proteins. Techniques from computer science are used in order to see if it is possible to predict protein mutations i.e. using artificial neural networks. The computer science perspective of this work would be to try optimizing the results from the neural networks. However, the focus of this thesis is primarily on the biological perspective and the performance of the computer science methods are secondary objective i.e. the primary interest is to show the existence of rules for protein mutations.</p><p>The method used in this thesis consists two neural networks. One network is used to predict the actual protein mutations and the other network is used to make a compressed representation of each amino acid. By using a compression network it is possible to make the prediction network much smaller (each amino acid is represented by 3 nodes instead of 22 nodes). The compression network is an auto associative network and the prediction network is a standard feed-forward network. The prediction network predicts a block of amino acids at a time and for comparison a sliding window technique has also been tested.</p><p>It is my belief that the results in this thesis indicate that there exists rules for protein mutations. However, the tests done in this thesis is only performed on a small portion of all proteins. Some protein families tested show really good results while other families are not as good. I believe that extended work using optimized neural networks would improve the predictions further.</p>

This thesis is concerned with the prediction of protein mutations using artificial neural networks. From the biological perspective it is of interest to investigate weather it is possible to find rules of mutation between evolutionary adjacent (or closely related) proteins. Techniques from computer science are used in order to see if it is possible to predict protein mutations i.e. using artificial neural networks. The computer science perspective of this work would be to try optimizing the results from the neural networks. However, the focus of this thesis is primarily on the biological perspective and the performance of the computer science methods are secondary objective i.e. the primary interest is to show the existence of rules for protein mutations. The method used in this thesis consists two neural networks. One network is used to predict the actual protein mutations and the other network is used to make a compressed representation of each amino acid. By using a compression network it is possible to make the prediction network much smaller (each amino acid is represented by 3 nodes instead of 22 nodes). The compression network is an auto associative network and the prediction network is a standard feed-forward network. The prediction network predicts a block of amino acids at a time and for comparison a sliding window technique has also been tested. It is my belief that the results in this thesis indicate that there exists rules for protein mutations. However, the tests done in this thesis is only performed on a small portion of all proteins. Some protein families tested show really good results while other families are not as good. I believe that extended work using optimized neural networks would improve the predictions further.

This thesis describes a new approach to the detection of protein stability change upon amino acid mutations. The main goal is to create a new meta-tool, which combines the outputs of eight well-established prediction tools and due to suitable method of consensus making, it is able to improve the overall prediction accuracy. The optimal strategy of combination of outputs of these tools is found by using a various number of machine learning methods. From all tested machine learning methods, KStar showed the highest prediction accuracy on the training dataset compiled from experimentally validated mutations originating from ProTherm database. Due to this reason, it is chosen as an optimal prediction technique. The general prediction abilities is validated on the testing dataset composed of multi-point amino acid mutations extracted also from ProTherm database. Since the multi-point mutations were not used for training any of integrated tools, we suppose that such comparison is objective. As a result, the developed meta-tool based on KStar technique improves the correlation coefficient about 0.130 on the training dataset and 0.239 on the testing dataset, respectively (the comparison is being made against the most succesful integrated tool). Based on the obtained results, it is possible to claim that machine learning methods are suitable technique for the problems from area of protein predictions.

La sensibilité et la profondeur de l'analyse protéomique sont limitées par les ions isobares et les interférences qui entravent l'identification des peptides de faible abondance. Lorsque nous analysons des échantillons de grande complexité, une séparation extensive de l'échantillon est souvent nécessaire pour étendre la couverture protéomique. Ces dernières années, la spectrométrie de mobilité ionique à forme d'onde asymétrique à haut champ (FAIMS) a gagné en popularité dans le domaine de la protéomique pour sa capacité à séparer les ions isobares, à améliorer la capacité de pic et la sensibilité de la spectrométrie de masse (MS). Nous rapportons ici l'intégration d'un appareil FAIMS Pro™ à un Q-Exactive HF™ ainsi qu'un spectromètre de masse Orbitrap Exploris 480™. Des expériences protéomiques sur des digestions d'extraits protéiques issues de cellules Hela à l'aide d'un spectromètre de masse avec FAIMS ont amélioré le rapport signal sur bruit (S/N) et réduit les ions interférents, ce qui a entraîné une augmentation du taux d'identification des peptides de plus de 42 %. FAIMS est également combiné avec le fractionnement ionique segmenté (SIFT), qui utilise tour à tour une fenêtre de 100 ~ 300 m/z au lieu de la large plage traditionnelle (700 ~ 800 m/z), augmentant ainsi la profondeur de la couverture protéomique tout en réduisant la proportion de spectres MS/MS chimériques de 50% à 27%. Dans l'analyse quantitative, nous démontrons l'application de FAIMS pour améliorer les mesures quantitatives lorsque le marquage peptidique isobare est utilisé. Par rapport aux expériences LC-MS/MS conventionnelles, la combinaison des expériences FAIMS et SIFT réalisées sur un modèle à deux protéomes a montré une amélioration de 65 % de la précision des mesures quantitatives. Les digestions tryptiques d'extraits protéiques de différentes lignées cellulaires du cancer colorectal ont été utilisées pour l'évaluation de stratégie combinée FAIMS et SIFT sur un spectromètre de masse Orbitrap Exploris 480™ offre un gain d'identification de 70 % par rapport à l'approche conventionnelle et combinée aux données transcriptomiques elle facilite l’identification de variants protéiques. / The sensitivity and depth of proteomic analysis in mass spectrometry (MS) is limited by isobaric ions and interferences that hinder the identification of low-abundance peptides. For high complexity samples, extensive separation is often required to expand proteomic coverage. In recent years, high-field asymmetric waveform ion mobility spectrometry (FAIMS) has gained popularity in the field of proteomics for its ability to resolve confounding ions, improve peak capacity, and sensitivity. This thesis presents the integration of a FAIMS Pro™ interface with electrical and gas embedded connections to a Q-Exactive HF™ as well as an Orbitrap Exploris 480™ mass spectrometer. Proteomic experiments on tryptic digests of HeLa cell line using a FAIMS integrated mass spectrometer improved signal-to-noise ratio (S/N) and reduced the occurrence of interfering ions. This enabled a 42% increase in peptide identification rate. Also, FAIMS was combined with segmented ion fractionation (SIFT), which in turn scans with windows of 100~300 m/z width instead of the traditional width (700~800 m/z), further increasing the depth of proteome coverage by a reducing from 50% to 27% in terms of MS/MS chimeric spectra numbers. The application of FAIMS gain improvement on quantitative measurements with TMT labeling method is presented. Compared to conventional LC-MS/MS tests, the combination of FAIMS and SIFT experiments showed a improvement by 65% in quantitative accuracy when performed on a human-yeast two-proteome model. As an application of the method, the tryptic digests from different colorectal cancer cell lines were used for the evaluation. FAIMS-SIFTcombined strategy on an Orbitrap Exploris 480™ mass spectrometer provides a 70% gain in identification compared to the conventional LC-MS/MS approach for the same sample amount and instrument time. This enhanced sensitivity facilitates single amino acid mutations confirmed by RNAseq analyses.