Yingyao Zhou 周颖耀
Active Recent Research
Publications in PubMed [Search]
Bioinformatics & Public Health
Metascape provides biologist-friendly gene list(s) analysis tools. We enthusiastically implement a suit of analysis features, so that all biomedical researcher can benefit from our decade-long bioinformatics experience. The tool has been published in Nature Communications and have been cited in over 3500 publications by May 2022.
I blog occasionally on related topics.
Machine Learning & Deep Learning,
We are fortunate to live in the middle of data science revolution. I enjoy learning the latest developments and help others enter these exciting fields. My understand of these subjects is summarized in my online machine learning book.
At work my team applied deep learning extensively to analyze biological images. We are currently developing an AI-based analysis platform based on CellPose, DeepCluster and single-cell population analysis.
In 2020 I led a winning team in a machine learning competition, where we aimed to predict the probability of success for Phase II drug candidates (published in Cell Patterns).
Cloud Computing & Big Data
Here is a short blog on Cloud computing. At work applications developed by my team run on the Kubernetes platform.
Recently the Metascape web site I championed was migrated onto AWS. We also provide a docker-container MSBio that enables bioinformaticians to run Metascape in batch mode.
Drug Discovery - Antimalarial Research
Research becomes most rewarding when it contributes directly to human health. In the past decade, we have gone a long way in finding the silver bullet for malaria. We took consecutive logical steps in dissecting P. falciparum parasite at biochemical level: we designed the first malaria gene chip to profile the parasite's life cycle transcription pattern [PubMed ID: 12893887], studied its transcription regulation mechanism [PubMed ID: 18257930], scrutinized the genome for best drug targets [PubMed ID: 16789840]; then combined genomics, proteomics and literature data to understand the parasite at systems level [PubMed ID: 18270564]. All these are enabling us to study parasite's host environment [PubMed ID: 18046333], and pave our way to use chemical genomics tools to find the potential cure [PubMed ID: 18579783]. We have applied machine learning to study resistance development in compounds [PubMed ID: 27301419] and further explore the resistome in P. falciparum [PubMed ID: 29326268]. Our latest efforts include the largest open-source research of malaria chemical leads [PubMed ID: 30523084].
Algorithm Development - Knowledge-based Optimization Analaysis (KOA)
With so many algorithms to choose from and so many parameters to tune, informaticians can be easily at lost. At the end of the day, the most popular algorithm is not always the best for your problem. The philosophy I have been relying on in nearly all my research works is: the algorithm that can best reproduce our existing knowledge has the best chance of rendering successful predictions, i.e., we need to optimize our algorithm and parameters based on prior knowledge.
Our first mathematical formulation of this idea is named Ontology-based Pattern Identification (OPI) algorithm, because it was first applied to gene expression-based gene function prediction problem [PubMed ID: 15531612]. The formulation was later improved in its application for better hit selection, known as Redundant siRNA Analaysis (RSA) [PubMed ID:17828270].
The many interesting applications of KOA has been summarized in a chapter of the book - Pharmaceutical Data Mining.
Copyright © Yingyao Zhou, 2006-2017