Probabilistic latent factor modelsfor transformative drug discovery
Başlık çevirisi mevcut değil.
- Tez No: 759871
- Danışmanlar: DR. ZİV BAR-JOSEPH
- Tez Türü: Doktora
- Konular: Biyoloji, Biyomühendislik, Biyoteknoloji, Biology, Bioengineering, Biotechnology
- Anahtar Kelimeler: Belirtilmemiş.
- Yıl: 2015
- Dil: İngilizce
- Üniversite: University of Pittsburgh
- Enstitü: Yurtdışı Enstitü
- Ana Bilim Dalı: Belirtilmemiş.
- Bilim Dalı: Belirtilmemiş.
- Sayfa Sayısı: 234
Özet
Özet yok.
Özet (Çeviri)
: The cost of discovering a new drug has doubled every 9 years since the 1950s. This can change by using machine learning to guide experimentation. The idea I have developed over the course of my PhD is that using latent factor modeling (LFM) of the drug-target interaction network, we can guide drug repurposable efforts to achieve transformative improvements. By better characterizing the drug-target interaction network, it is possible to use currently approved drugs to achieve therapies for diseases that currently are not optimally treated. These drugs might be directly used through repurposing, or they can serve as a starting point for new drug discovery efforts where they are optimized through medicinal chemistry methods. To achieve this goal, I have developed LFM-based techniques applicable to existing databases of drug-target interaction networks. Specifically, I have started out by establishing that probabilistic matrix factorization (PMF; one type of LFM algorithm) can be used as descriptors by showing they capture therapeutic function similarities that state-of-the-art 3D chemical similarity methods could not capture. Then I have shown that PMF can effectively predict unknown drug-target interactions. Furthermore, I have used newly developed computational techniques for discovering repurposable drugs for two diseases, α1 antitrypsin (1-AT) deficiency (ATD) and Huntington's disease (HD) leading to successful discoveries in both. For ATD, two sets of data generated by the David Perlmutter and Gary Silverman laboratories have been used as input to deduce potential targets and repurposable drugs: (i) a high throughput screening data from a genomewide RNAi knockdown in a C. elegans model for studying ATZ (Z-allele of 1-AT), and (ii) data from Prestwick library screen for the same model. We have predicted that the antidiabetic drug glibenclamide would be beneficial against ATZ aggregation, and data collected to date in Mus musculus models are promising. We have worked on HD with the Robert Friedlander lab, v by examining the potential drugs and implicated pathways for 15 neuroprotective (repurposable) drugs that they have identified in a two-stage screening study. Based on LFM-based analysis of the targets of these drugs, we have developed a number of hypotheses to be tested. Among them, the antihypertensive drug sodium nitroprusside appears to be effective against HD based on neuronal cell death inhibition experiments that were conducted at the University of Pittsburgh Drug Discovery Institute as well as the Friedlander lab. Finally, we have built a web server, named BalestraWeb, for facilitating the use of PMF in repurposable drug identification by the broader community. BalestraWeb enables users to extract information on known and potential targets (or drugs) for any approved drug (or target), simply by entering the name of the query drug (or target). I have also laid out the framework for developing an integrated resource for quantitative systems pharmacology, Balestra toolkit (BalestraTK), which would take advantage of existing databases such as STITCH, UniProt, and PubChem. Collectively, our results provide firm evidence for the potential utility of machine learning techniques for assisting in drug discovery.
Benzer Tezler
- Learning mental states from biosignal
Başlık çevirisi yok
MELİH KANDEMİR
Doktora
İngilizce
2013
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolAalto University (Aalto Yliopisto)PROF. SAMUEL KASKI
DR. ARTO KLAMI
- Probabilistic tensor factorization for link prediction
Bağlantı tamini için olasılıksal tensör ayrışımı
BEYZA ERMİŞ
Yüksek Lisans
İngilizce
2012
Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolBoğaziçi ÜniversitesiBilgisayar Mühendisliği Ana Bilim Dalı
DOÇ. ALİ TAYLAN CEMGİL
- Gemi radar sistemlerinde uzaktan kod yürütme (RCE) tehditlerini azaltmaya yönelik siber güvenlik önlemlerinin bayes ağı metoduyla değerlendirilmesi
Evaluation of cybersecurity measures to mitigate remote code execution (RCE) threats in shipboard radar systems using the bayesian network method
EMRE DÜZENLİ
Yüksek Lisans
Türkçe
2024
Deniz Bilimleriİstanbul Teknik ÜniversitesiDeniz Ulaştırma Mühendisliği Ana Bilim Dalı
DOÇ. DR. PELİN BOLAT
- En uygun hayat sigortası poliçesi seçimini sağlayan bir karar modeli
A Decision model for selecting the optimum insurance policy
H.BÜLENT CERİT
Yüksek Lisans
Türkçe
1995
Endüstri ve Endüstri Mühendisliğiİstanbul Teknik ÜniversitesiPROF.DR. RAMAZAN EVREN