東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Learning Similarity Functions on Per...

Smith, Ian,

Linked to FindBook

Google Book

Amazon

博客來

Learning Similarity Functions on Perturbational Biological Data /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Learning Similarity Functions on Perturbational Biological Data // Ian Smith.
Author:	Smith, Ian,
Description:	1 electronic resource (170 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
Contained By:	Dissertations Abstracts International86-05B.
Subject:	Bioinformatics. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30992784
ISBN:	9798342742733

Learning Similarity Functions on Perturbational Biological Data /
Smith, Ian,

Learning Similarity Functions on Perturbational Biological Data /Ian Smith. - 1 electronic resource (170 pages)

Source: Dissertations Abstracts International, Volume: 86-05, Section: B.

Precision cancer medicine provides an avenue for optimizing patient care based on the therapies that will be most effective to treat each cancer patient's specific disease. However, there are two major challenges in precision medicine: (1) identifying predictive biomarkers to match patients to therapeutics and (2) discovering new therapeutic agents to increase the arsenal of oncology treatments. Existing cancer treatments are often unsuccessful because of these challenges. Massive, high-throughput pharmacogenomic and phenotypic data resources have been and continue to be generated to identify biomarkers and characterize potential new therapeutics, but these datasets produce inconsistent results and have substantial statistical and biological noise that obscures signal.In this thesis, I introduce methods for overcoming limitations of pharmacogenomic and phenotypic analyses. First, I introduce and investigate the robust Concordance Index (rCI), a semi-parametric extension of the concordance index designed to overcome noise in drug sensitivity screens. The rCI demonstrates theoretical performance improvements and is better powered than some existing methods, but is less powered than Pearson's correlation for detecting monotonic associations. Turning to phenotypic analysis, I next characterize a previously unknown similarity bias that arises from consensus signatures on perturbational datasets caused by erroneous assumptions about the underlying data distribution. Third, I derive the relationship between the distribution of data and the distribution of cosine similarity on that data. While this relationship provides fundamental insight into similarities computed on datasets, the main contribution is identifying how transformations or representations of phenotypic data can affect the power to find related perturbations. Finally, I introduce perturbational metric learning (PeML), a weakly supervised method to learn a similarity metric directly from perturbational datasets. PeML, when applied to the L1000 Connectivity Map and Cell Painting morphological datasets, shows substantial improvement at identifying compounds with the same mechanism of action.Together, these contributions provide new insights, methods, and tools that can advance pharmacogenomic and phenotypic analysis, and ultimately improve patient care.

English

ISBN: 9798342742733Subjects--Topical Terms:

553671
Bioinformatics.
Subjects--Index Terms:

Computational biology

Learning Similarity Functions on Perturbational Biological Data /
LDR:03854nmm a22004573i 4500 001 2434064
005 20250905091337.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798342742733
035 $a (MiAaPQD)AAI30992784
035 $a AAI30992784
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Smith, Ian, $e author. $0 (orcid)0000-0003-3957-4397 $3 3823918
245 1 0 $a Learning Similarity Functions on Perturbational Biological Data / $c Ian Smith.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (170 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-05, Section: B.
500 $a Advisors: Haibe-Kains, Benjamin Committee members: Goldenberg, Anna; Lupien, Mathieu.
502 $b Ph.D. $c University of Toronto (Canada) $d 2024.
520 $a Precision cancer medicine provides an avenue for optimizing patient care based on the therapies that will be most effective to treat each cancer patient's specific disease. However, there are two major challenges in precision medicine: (1) identifying predictive biomarkers to match patients to therapeutics and (2) discovering new therapeutic agents to increase the arsenal of oncology treatments. Existing cancer treatments are often unsuccessful because of these challenges. Massive, high-throughput pharmacogenomic and phenotypic data resources have been and continue to be generated to identify biomarkers and characterize potential new therapeutics, but these datasets produce inconsistent results and have substantial statistical and biological noise that obscures signal.In this thesis, I introduce methods for overcoming limitations of pharmacogenomic and phenotypic analyses. First, I introduce and investigate the robust Concordance Index (rCI), a semi-parametric extension of the concordance index designed to overcome noise in drug sensitivity screens. The rCI demonstrates theoretical performance improvements and is better powered than some existing methods, but is less powered than Pearson's correlation for detecting monotonic associations. Turning to phenotypic analysis, I next characterize a previously unknown similarity bias that arises from consensus signatures on perturbational datasets caused by erroneous assumptions about the underlying data distribution. Third, I derive the relationship between the distribution of data and the distribution of cosine similarity on that data. While this relationship provides fundamental insight into similarities computed on datasets, the main contribution is identifying how transformations or representations of phenotypic data can affect the power to find related perturbations. Finally, I introduce perturbational metric learning (PeML), a weakly supervised method to learn a similarity metric directly from perturbational datasets. PeML, when applied to the L1000 Connectivity Map and Cell Painting morphological datasets, shows substantial improvement at identifying compounds with the same mechanism of action.Together, these contributions provide new insights, methods, and tools that can advance pharmacogenomic and phenotypic analysis, and ultimately improve patient care.
546 $a English
590 $a School code: 0779
650 4 $a Bioinformatics. $3 553671
650 4 $a Statistics. $3 517247
650 4 $a Biophysics. $3 518360
650 4 $a Medicine. $3 641104
653 $a Computational biology
653 $a Cosine similarity
653 $a Perturbational data
653 $a Pharmacogenomics
653 $a Similarity metrics
653 $a Weakly supervised learning
690 $a 0715
690 $a 0463
690 $a 0786
690 $a 0564
710 2 $a University of Toronto (Canada). $b Medical Biophysics. $e degree granting institution. $3 3823919
720 1 $a Haibe-Kains, Benjamin $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-05B.
790 $a 0779
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30992784