December, 2020

In-born genomes impact  on  cancer risk, the work has been published in Science Advances , 2020 (online). 

Related News from Yahoo, MSN, and Science Magazine 

October, 2019

A new machine learning algorithm for more accurately predicting breast cancer recurrence and metastasis by sequencing of patient's blood  samples (a non-invasive genomic test)

In the past 2 decades, gene expression profiles or DNA sequencing data  (i.e., tumor circulating DNAs) of tumours have been used for predicting cancer patient prognosis (recurrence and metastasis) to guide in making treatment decisions. For example, Oncotype DX uses gene expression profiles of breast tumor samples to predict who has a higher chance to recur in the future so that she should be treated aggressively.

Based on previously proposed 'Cancer Hallmark Network Framework' (Seminars in Cancer Biology, 2015;  arXiv, 2014), Wang Lab has developed a machine learning algorithm which enabled to predict breast cancer recurrence and prognosis by whole-exome sequencing of patient's blood samples (germline variants) (npj [Nature Partner Journal]  Precision Oncology, 3:28, 2019; paper onlinebioRxiv, 2018a;  bioRvix, 2018b)The prediction is better than that of the widely used  Oncotype DX. Theoretically, it further validates the concept that germline genomes play a critical role in tumorigenesis and metastasis (Seminars in Cancer Biology, 2015; arXiv, 2014; JAMA Network Open, 2019), while practically, it provides a non-invasive means for predicting cancer prognosis.  

--News from the GenomeWeb (news here)    --News from the LabMedica (news here)  --Blog at Nature Research Cancer (blog

September, 2019

Germline pathogenic genes in NK cells shape tumorigenesis, metastasis, tumour-infiltrating lymphocytes (TIL) and immunotherapy 

The first study about the effects of germline pathogenic genes in immune system on tumorigenesis and tumor immune microenvironment (TIME) subtypes showed that the number of  germline pathogenic genes  in NK cells are negatively correlated with abundance of TILs, rich-TIME subtype, immunotherapy response, disease-free survival, but positively correlated with cancer risk. See JAMA Network Open (paper online) 

--Blog at Nature Research Cancer (blog

October, 2015

A clinical question debated for more than 20 years has been resolved by developing machine learning algorithms  

"Treat or not treat" for colon cancer stage 2 patients, an issue which has been debated for more than 20 years, has been successfully resolved by developing a new machine learning algorithm and combinatory cancer hallmark-based gene signature sets. See JAMA Oncology, 2016 (paper online) and the news report from Reuters (PDF). Our previous developed machine learning algorithm (Nature Communications, 2010, paper online) has been used in this study.  

May, 2014

The Cancer Hallmark Network Framework has been published

The Hallmarks of Cancer represent the most important understanding of cancer in the past of 50 years. However, these hallmarks have been largely descriptive. This framework enables the cancer hallmarks to be represented, quantified and further modeled computationally  using molecular networks. The quantified models allow examining the collective effects of genomic alterations on cancer hallmark networks for predicting clinical outcomes. See Seminars in Cancer Biology (paper online). 

March, 2014/2017

The work of the cancer network motifs has been featured in a college textbook, Genetics (2014, 2017)

The pioneering work of the cancer network motifs with a series of reports to show the molecular mechanisms of cancer progression and clinical features in the context of network motifs (PDF) has been featured in the textbook, Genetics. The textbook (published by McGraw-Hill Ryerson, 2014 and 2017) is edited by a Nobel Laureate Dr. Hartwell and the father of systems biology, Dr. Hood. The original research work were published in Molecular Systems Biology, Genome Research and other journals during 2007-2012. 

July, 2010

A new machine learning algorithm for identifying robust  and more accurate  cancer biomarkers

Due to tumor heterogeneity, most of the comics-based cancer biomarkers are not robust - meaning that they often lost their predictive power in new clinical samples. To overcome this problem, a new machine learning algorithm, MSS (Multiple Survival Screening) has been developed by double-resampling of samples and genes, and incorporating cancer hallmark concepts. The excellent performance of the MSS-derived biomarkers have been successfully test in all the available data cohorts. The MSS algorithm has been published in Nature Communications, 2010, paper online, and News.  

December, 2007

Uncovering cancer network motifs by constructing a map of human cancer signalling

The pioneering work of the cancer network motifs has been published in Molecular Systems Biology, 2007  (paper online), and News.  

September, 2006

The pioneering work of the microRNA networks has been published, which opens a new research area 

The pioneering work of the microRNA networks has been published. Up to 2018, >7,000 papers in this research area (miRNA/lncRNA networks) have been published in Molecular Systems Biology, 2006  (paper online).   



Genomic data analysis (Whole-exome, RNA-seq, Methylation, CNV, tumor only): see here (pdf), website (html)

Single-cell genomic data analysis: see here

- SingleCellNet (classify single cells across species and platforms): see here

Omic analysis tools: see here 

-Comments on RNA-seq data normalization: (1) comparing of assumptions and methods for normalization (here);  



Basic Biology (video  from MIT) - link (video)

Computational Genomics - link (video)

Basic Deep Learning in Genomics - link (PPT and Videos)

Genomic regulation - link (video), link2 (video)Single-cell RNA-seq and analysis - link (video)

Molecular Networks - link (video)