Evolutionary Algorithms for Improving De Novo Peptide Sequencing

<p>De novo peptide sequencing algorithms have been developed for peptide identification in proteomics from tandem mass spectra (MS/MS), which can be used to identify and discover novel peptides and proteins that do not have a database available. Despite improvements in MS instrumentation and de novo sequencing methods, a significant number of CID MS/MS spectra still remain unassigned with the current algorithms, often leading to low confidence of peptide assignments to the spectra. Moreover, current algorithms often fail to construct the completely matched sequences, and produce partial matches. Therefore, identification of full-length peptides remains challenging. Another major challenge is the existence of noise in MS/MS spectra which makes the data highly imbalanced. Also missing peaks, caused by incomplete MS fragmentation makes it more difficult to infer a full-length peptide sequence. In addition, the large search space of all possible amino acid sequences for each spectrum leads to a high false discovery rate. This thesis focuses on improving the performance of current methods by developing new algorithms corresponding to three steps of preprocessing, sequence optimisation and post-processing using machine learning for more comprehensive interrogation of MS/MS datasets. From the machine learning point of view, the three steps can be addressed by solving different tasks such as classification, optimisation, and symbolic regression. Since Evolutionary Algorithms (EAs), as effective global search techniques, have shown promising results in solving these problems, this thesis investigates the capability of EAs in improving the de novo peptide sequencing. In the preprocessing step, this thesis proposes an effective GP-based method for classification of signal and noise peaks in highly imbalanced MS/MS spectra with the purpose of having a positive influence on the reliability of the peptide identification. The results show that the proposed algorithm is the most stable classification method across various noise ratios, outperforming six other benchmark classification algorithms. The experimental results show a significant improvement in high confidence peptide assignments to MS/MS spectra when the data is preprocessed by the proposed GP method. Moreover, the first multi-objective GP approach for classification of peaks in MS/MS data, aiming at maximising the accuracy of the minority class (signal peaks) and the accuracy of the majority class (noise peaks) is also proposed in this thesis. The results show that the multi-objective GP method outperforms the single objective GP algorithm and a popular multi-objective approach in terms of retaining more signal peaks and removing more noise peaks. The multi-objective GP approach significantly improved the reliability of peptide identification. This thesis proposes a GA-based method to solve the complex optimisation task of de novo peptide sequencing, aiming at constructing full-length sequences. The proposed GA method benefits the GA capability of searching a large search space of potential amino acid sequences to find the most likely full-length sequence. The experimental results show that the proposed method outperforms the most commonly used de novo sequencing method at both amino acid level and peptide level. This thesis also proposes a novel method for re-scoring and re-ranking the peptide spectrum matches (PSMs) from the result of de novo peptide sequencing, aiming at minimising the false discovery rate as a post-processing approach. The proposed GP method evolves the computer programs to perform regression and classification simultaneously in order to generate an effective scoring function for finding the correct PSMs from many incorrect ones. The results show that the new GP-based PSM scoring function significantly improves the identification of full-length peptides when it is used to post-process the de novo sequencing results.</p>

Evolutionary Algorithms for Improving De Novo Peptide Sequencing

<p>De novo peptide sequencing algorithms have been developed for peptide identification in proteomics from tandem mass spectra (MS/MS), which can be used to identify and discover novel peptides and proteins that do not have a database available. Despite improvements in MS instrumentation and de novo sequencing methods, a significant number of CID MS/MS spectra still remain unassigned with the current algorithms, often leading to low confidence of peptide assignments to the spectra. Moreover, current algorithms often fail to construct the completely matched sequences, and produce partial matches. Therefore, identification of full-length peptides remains challenging. Another major challenge is the existence of noise in MS/MS spectra which makes the data highly imbalanced. Also missing peaks, caused by incomplete MS fragmentation makes it more difficult to infer a full-length peptide sequence. In addition, the large search space of all possible amino acid sequences for each spectrum leads to a high false discovery rate. This thesis focuses on improving the performance of current methods by developing new algorithms corresponding to three steps of preprocessing, sequence optimisation and post-processing using machine learning for more comprehensive interrogation of MS/MS datasets. From the machine learning point of view, the three steps can be addressed by solving different tasks such as classification, optimisation, and symbolic regression. Since Evolutionary Algorithms (EAs), as effective global search techniques, have shown promising results in solving these problems, this thesis investigates the capability of EAs in improving the de novo peptide sequencing. In the preprocessing step, this thesis proposes an effective GP-based method for classification of signal and noise peaks in highly imbalanced MS/MS spectra with the purpose of having a positive influence on the reliability of the peptide identification. The results show that the proposed algorithm is the most stable classification method across various noise ratios, outperforming six other benchmark classification algorithms. The experimental results show a significant improvement in high confidence peptide assignments to MS/MS spectra when the data is preprocessed by the proposed GP method. Moreover, the first multi-objective GP approach for classification of peaks in MS/MS data, aiming at maximising the accuracy of the minority class (signal peaks) and the accuracy of the majority class (noise peaks) is also proposed in this thesis. The results show that the multi-objective GP method outperforms the single objective GP algorithm and a popular multi-objective approach in terms of retaining more signal peaks and removing more noise peaks. The multi-objective GP approach significantly improved the reliability of peptide identification. This thesis proposes a GA-based method to solve the complex optimisation task of de novo peptide sequencing, aiming at constructing full-length sequences. The proposed GA method benefits the GA capability of searching a large search space of potential amino acid sequences to find the most likely full-length sequence. The experimental results show that the proposed method outperforms the most commonly used de novo sequencing method at both amino acid level and peptide level. This thesis also proposes a novel method for re-scoring and re-ranking the peptide spectrum matches (PSMs) from the result of de novo peptide sequencing, aiming at minimising the false discovery rate as a post-processing approach. The proposed GP method evolves the computer programs to perform regression and classification simultaneously in order to generate an effective scoring function for finding the correct PSMs from many incorrect ones. The results show that the new GP-based PSM scoring function significantly improves the identification of full-length peptides when it is used to post-process the de novo sequencing results.</p>

Protein Identification Strategies for the Greenshell Mussel Perna canaliculus

<p>The greenshell mussel Perna canaliculus is considered to be a suitable biomonitor for heavy metal pollution. This is due to their ability to accumulate and tolerate heavy metals in their tissues. These characteristics make them useful for identifying protein biomarkers of heavy metal pollution, as well as proteins associated with heavy metal detoxification and homeostasis. However, the identification of such proteins is restricted by the greenshell mussel being poorly represented in sequence databases. Several strategies have previously been used to identify proteins in unsequenced species, but only one of these strategies has been applied to the greenshell mussel. The objective of this thesis was to examine different protein identification strategies using a combined two-dimensional gel electrophoresis and MALDI-TOF/TOF mass spectrometry approach. The protein identification strategies used include a Mascot database search, as well as de novo sequencing approaches using PEAKS DB and SPIDER homology searches. In total, 155 protein spots were excised and a total of 68 identified. Fifty-six proteins were identified using a Mascot search against the Mollusca, NCBInr and Invertebrate EST database, with seven single-peptide identifications. De novo sequencing strategies identified additional proteins, with two from a PEAKS DB search and 10 from an error-tolerant SPIDER homology search. The most noticeable protein groups identified were cytoskeletal proteins, stress response proteins and those involved in protein biosynthesis. Actin and tubulin made up the bulk of the identifications, accounting for 39% of all proteins identified. This multifaceted approach was shown to be useful for identifying proteins in the greenshell mussel Perna canaliculus. Mascot and PEAKS DB performed equally well, while the error-tolerant functionality of SPIDER was useful for identifying additional proteins. A subsequent search against the Invertebrate EST database was also found to be useful for identifying additional proteins. Despite this, more than half of all proteins remained unidentified. Most of these proteins either failed to produce good quality MS spectra or did not find a match to a sequence in the database. Future research should first focus on obtaining quality MS spectra for all proteins concerned and then examine other strategies that may be more suitable for identifying proteins for species with poor representation in sequence databases.</p>

Protein Identification Strategies for the Greenshell Mussel Perna canaliculus

<p>The greenshell mussel Perna canaliculus is considered to be a suitable biomonitor for heavy metal pollution. This is due to their ability to accumulate and tolerate heavy metals in their tissues. These characteristics make them useful for identifying protein biomarkers of heavy metal pollution, as well as proteins associated with heavy metal detoxification and homeostasis. However, the identification of such proteins is restricted by the greenshell mussel being poorly represented in sequence databases. Several strategies have previously been used to identify proteins in unsequenced species, but only one of these strategies has been applied to the greenshell mussel. The objective of this thesis was to examine different protein identification strategies using a combined two-dimensional gel electrophoresis and MALDI-TOF/TOF mass spectrometry approach. The protein identification strategies used include a Mascot database search, as well as de novo sequencing approaches using PEAKS DB and SPIDER homology searches. In total, 155 protein spots were excised and a total of 68 identified. Fifty-six proteins were identified using a Mascot search against the Mollusca, NCBInr and Invertebrate EST database, with seven single-peptide identifications. De novo sequencing strategies identified additional proteins, with two from a PEAKS DB search and 10 from an error-tolerant SPIDER homology search. The most noticeable protein groups identified were cytoskeletal proteins, stress response proteins and those involved in protein biosynthesis. Actin and tubulin made up the bulk of the identifications, accounting for 39% of all proteins identified. This multifaceted approach was shown to be useful for identifying proteins in the greenshell mussel Perna canaliculus. Mascot and PEAKS DB performed equally well, while the error-tolerant functionality of SPIDER was useful for identifying additional proteins. A subsequent search against the Invertebrate EST database was also found to be useful for identifying additional proteins. Despite this, more than half of all proteins remained unidentified. Most of these proteins either failed to produce good quality MS spectra or did not find a match to a sequence in the database. Future research should first focus on obtaining quality MS spectra for all proteins concerned and then examine other strategies that may be more suitable for identifying proteins for species with poor representation in sequence databases.</p>

Spectrum graph-based de-novo sequencing algorithm MaxNovo achieves high peptide identification rates in collisional dissociation MS/MS spectra

We describe MaxNovo, a novel spectrum graph-based peptide de-novo sequencing algorithm integrated into the MaxQuant software. It identifies complete sequences of peptides as well as sequence tags that are incomplete at one or both of the peptide termini. MaxNovo searches for the highest-scoring path in a directed acyclic graph representing the MS/MS spectrum with peaks as nodes and edges as potential sequence constituents consisting of single amino acids or pairs. The raw score is a sum of node and edge weights, plus several reward scores, for instance, for complementary ions or protease compatibility. For search-engine identified peptides, it correlates well with the Andromeda search engine score. We use a particular score normalization and the score difference between the first and second-best solution to define a combined score that integrates all available information. To evaluate its performance, we use a human cell line dataset and take as ground truth all Andromeda-identified MS/MS spectra with an Andromeda score of at least 100. MaxNovo outperforms other software in particular in the high-sensitivity range of precision-coverage plots. We also identify incomplete sequence tags and study their statistical properties. Next, we apply MaxNovo to ion mobility-coupled time of flight data. Here we achieve excellent performance as well, except for potential swaps of the two amino acids closest to the C-terminus, which are not well resolved due to the low end of the mass range in MS/MS spectra in this dataset. We demonstrate the applicability of MaxNovo to palaeoproteomics samples with a Late Pleistocene hominin proteome dataset that was generated using three proteases. Interestingly, we did not use any machine learning in the construction of MaxNovo, but implemented expert domain knowledge directly in the definition of the score. Yet, it performs as good as or better than the leading deep learning-based algorithm.