Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


41629.pdf (8.71 MB)

ETD Abstract Container

Abstract Header

Learning the Mechanisms of Action of Chemical Perturbagens from their Transcriptional Signatures

Zhang, Lixia
http://orcid.org/0000-0003-4094-0729
http://rave.ohiolink.edu/etdc/view?acc_num=ucin1649769278503312

Abstract Details

2022, PhD, University of Cincinnati, Medicine: Biostatistics (Environmental Health).
Mechanisms of action (MoA) of bioactive compounds represents the direct effects that one compound perturbagen has on the activity of specific genes, proteins, and pathways in a cell. As bioactive compounds, clinical drugs treat human diseases by modulating the activity of biological systems. Identification of MoA can help discover new candidates for known therapeutic targets, and unveil pharmacological actions for those approved drugs with unknown targets, which is of vital importance in medical and pharmaceutical research. Genome-wide expression profile, which shows transcriptional responses to bioactive compound perturbagen in human cell lines, delivers a good approach for MoA analysis. The Library of Integrated Network-based Cellular Signatures (LINCS) L1000 data is a transcriptome resource that stores a huge number of gene expression profiles, which is very valuable for understanding the relationships between compounds, genes and diseases as well as advancing the development of new drug therapies. Thus, we are highly motivated to develop novel computational methods by exploring this resource. In this dissertation, we provided novel methods to elucidate MoA of biological compounds by using the large-scale LINCS transcriptomic data. Briefly, we firstly compared the methods of measuring concordances between a Gene Expression Profiles of Interest (GEPI) which is a list of differential gene expressions for a biological state perturbed by a query compound and reference LINCS signatures. Then the calculated concordance measures from the best method were subjected to compounds detection with Random Set (RS) Enrichment Analysis. To predict MoA, a multilevel hierarchical MoA-Compound-Signature data were constructed by aggregating Compound-MoA from public data with the LINCS-derived Compound–Signature. Next, similar to compound enrichment, concordance measures between GEPI and reference signatures were calculated with the best-benchmarked method. Both MoA-Compound-Signature data and the calculated concordance measures were then subjected to MoA detection that was achieved by a Nested Linear Model (NLM). Lastly, the novel methods used for compounds and MoA detection for a single GEPI were expanded to all GEPIs from the LINCS trt_cp signatures. Overall, our methods showed a good performance (sensitivity and specificity) in identifying compounds and MoA for GEPIs, with the median AUC values of 0.75 and 0.6 respectively. Moreover, the AUC values from our novel approach were compared with those of the clue.io query results, which showed that our results are comparable to query results from clue.io and further validated the usefulness of our new methodology in compound and MoA detection. The originality of the study lies in new method flows for enriching compounds and MoA respectively were developed. We demonstrated the usefulness of the new method flows for the detection of compounds and MoA in terms of prediction accuracy, interpretability, and large-scale applicability. Therefore, this novel approach is anticipated to advance deep understandings of MoA, which is vital for drug repositioning of existing drugs, identifying novel drugs and predicting drug side effects during drug discovery.
Mario Medvedovic, Ph.D. (Committee Member)
Siva Sivaganesan, Ph.D. (Committee Member)
Liang Niu, Ph.D. (Committee Member)
Anil Jegga, DVM MRes (Committee Member)
144 p.
Biostatistics

Recommended Citations

Citations

  • Zhang, L. (2022). Learning the Mechanisms of Action of Chemical Perturbagens from their Transcriptional Signatures [Doctoral dissertation, University of Cincinnati]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1649769278503312

    APA Style (7th edition)

  • Zhang, Lixia. Learning the Mechanisms of Action of Chemical Perturbagens from their Transcriptional Signatures. 2022. University of Cincinnati, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=ucin1649769278503312.

    MLA Style (8th edition)

  • Zhang, Lixia. "Learning the Mechanisms of Action of Chemical Perturbagens from their Transcriptional Signatures." Doctoral dissertation, University of Cincinnati, 2022. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1649769278503312

    Chicago Manual of Style (17th edition)

ucin1649769278503312
101
© 2022, all rights reserved.
This open access ETD is published by University of Cincinnati and OhioLINK.