Combination and Tuning of MS/MS Spectra Scoring Functions

Viktoria Dorfer

Combination and Tuning of MS/MS Spectra Scoring Functions

Publikation: Typen von Abschlussarbeiten › Masterarbeit / Diplomarbeit

Abstract

Tandem Massen-Spektrometrie ist eine weit verbreitete und mächtige Methode zur Identizierung von Peptidsequenzen. Viele verschiedene Bewertungsfunktionen, die für eine korrekte Identikation benötigt werden, wurden schon in verschiedensten Journalen veröffentlicht, basierend auf unterschiedlichen Ansätzen. Jedoch kann das momentane Wissen über die wichtigen Faktoren in der Tandem Massen-Spektrometrie nicht immer in die Bewertungsfunktionen miteinbezogen werden. Diese Diplomarbeit stellt verschiedene Verbesserungsansätze vor, um bestehende Bewertungsfunktionen durch Einbeziehung einiger dieser Erkenntnisse zu optimieren. Es wird gezeigt, dass nur geringe Änderungen in diesen Bewertungsfunktionen nötig sind, um eine signikante Verbesserung zu erzielen. Zwei unterschiedliche Modelle zur Verbesserung der Ergebnisse werden in diesem Dokument behandelt. Bewertungsfunktionen basierend auf dem naïven Bayesschen Klassikator werden verbessert und kombiniert, um Optimierungen zu erhalten. Noch bessere Ergebnisse werden durch maschinelles Lernen erzielt, wo nun zu hoffen ist, dass eine neue Bewertungsfunktion, die signikant besser ist als Bestehende, entworfen werden kann.

Originalsprache	Englisch
Publikationsstatus	Veröffentlicht - 2007

Schlagwörter

MS-MS
Scoring Function
Tandem Mass Spectrometry

Zitieren

@mastersthesis{72708b9ab47d4546a4bf8a27c4c6915a,

title = "Combination and Tuning of MS/MS Spectra Scoring Functions",

abstract = "Tandem mass spectrometry is a powerful and commonly used method to identify protein sequences. Many different scoring functions, needed for a correct identication, have already been proposed in various journals, being based on most diverse approaches. Often, the current knowledge about the important issues in tandem mass spectrometry cannot be considered in the scoring functions. This thesis proposes different, simple approaches to tune existing scoring functions by using some of these ndings. It is shown that only small changes in the scoring functions are required to improve the results signicantly. Two main approaches for improving the results are dealt with in this document. Scoring functions based on na{\"i}ve Bayesian classiers are tuned and also combined to yield optimization. Even better prospects are achieved with machine learning methods, where it is hoped that a new scoring function, signi cantly better than the existing ones, can be designed.",

keywords = "MS-MS, Scoring Function, Tandem Mass Spectrometry, MS-MS, Scoring Function, Tandem Mass Spectrometry",

author = "Viktoria Dorfer",

year = "2007",

language = "English",

}

TY - THES

T1 - Combination and Tuning of MS/MS Spectra Scoring Functions

AU - Dorfer, Viktoria

PY - 2007

Y1 - 2007

N2 - Tandem mass spectrometry is a powerful and commonly used method to identify protein sequences. Many different scoring functions, needed for a correct identication, have already been proposed in various journals, being based on most diverse approaches. Often, the current knowledge about the important issues in tandem mass spectrometry cannot be considered in the scoring functions. This thesis proposes different, simple approaches to tune existing scoring functions by using some of these ndings. It is shown that only small changes in the scoring functions are required to improve the results signicantly. Two main approaches for improving the results are dealt with in this document. Scoring functions based on naïve Bayesian classiers are tuned and also combined to yield optimization. Even better prospects are achieved with machine learning methods, where it is hoped that a new scoring function, signi cantly better than the existing ones, can be designed.

AB - Tandem mass spectrometry is a powerful and commonly used method to identify protein sequences. Many different scoring functions, needed for a correct identication, have already been proposed in various journals, being based on most diverse approaches. Often, the current knowledge about the important issues in tandem mass spectrometry cannot be considered in the scoring functions. This thesis proposes different, simple approaches to tune existing scoring functions by using some of these ndings. It is shown that only small changes in the scoring functions are required to improve the results signicantly. Two main approaches for improving the results are dealt with in this document. Scoring functions based on naïve Bayesian classiers are tuned and also combined to yield optimization. Even better prospects are achieved with machine learning methods, where it is hoped that a new scoring function, signi cantly better than the existing ones, can be designed.

KW - MS-MS

KW - Scoring Function

KW - Tandem Mass Spectrometry

KW - MS-MS

KW - Scoring Function

KW - Tandem Mass Spectrometry

M3 - Master's Thesis / Diploma Thesis

ER -

Combination and Tuning of MS/MS Spectra Scoring Functions

Abstract

Schlagwörter

Fingerprint

Zitieren