DBDS in the News

AI can be sexist and racist — it’s time to make it fair

Computer scientists must identify sources of bias, de-bias training data and develop artificial-intelligence algorithms that are robust to skews in the data, argue James Zou (Stanford DBDS) and Londa Schiebinger.


Special Seminar by Dr. Mohit Kaushal on June 25, 2018!

Overview of the healthcare ecosystem: 101 of Macroeconomics, Health Policy, Market Reactions to Policy, and New Opportunities for Startups

Mohit Kaushal, MD (Adjunct Professor of Biomedical Data Science, Stanford)

Monday, June 25th 2:00 – 3:30pm in MSOB x303

Summary 

Almost no country anywhere in the world can afford their healthcare. Starbucks spends more money on healthcare than their coffee. General Motors spends more money on healthcare than steel for their cars. The macroeconomic effects of healthcare costs, along with significant sociodemographic shifts, impact every part of our society. In the United States, there have been twin massive policy responses (health technology and health payment reform), which are now being coupled with unprecedented venture and industry activity. 

This seminar will provide an overview of the US healthcare system, policy,and real world- and industry- based data opportunities within the healthcare system.

New publication in PLOS Genetics from the Rivas Lab!

DBDS faculty member Dr. Manuel Rivas's work, entitled "Insights into the genetic epidemiology of Crohn's and rare diseases in the Ashkenazi Jewish population," has been published today in PLOS Genetics

The authors summarize the study as follows: "The Ashkenazim are a people with ancestry in northern-European Jewish groups. A founder effect caused a bottleneck in this population approximately one thousand years ago, resulting in a group of enriched alleles in their genetic makeup. A higher documented prevalence of Crohn’s Disease in the Ashkenazim indicates that some enriched alleles may confer risk of having this disease. By studying which genes are enriched, and which of these contribute to Crohn’s Disease risk, we are better able to understand the genetic architecture of the affected population, and of the disease itself. Further, we are able to develop a resource containing tables of significantly enriched alleles that are known or suspected to contribute to other disease."


UK Biobank data opens up window into genetics of disease

"A trove of genetic records from the UK Biobank was unleashed last July, and after mining the immense data set, scientists at Stanford have found strong evidence for 27 direct links between specific genetic mutations and a variety of human diseases.

Some of these mutations have never been associated with disease before, and some even seem to confer protection against disease, making them prime drug candidates.

The team's work, published in Nature Communications, focused in on a particular type of genetic oddity that halts proper protein formation before it's complete. It's called a protein-truncating variant, or a PTV. Manuel Rivas, PhD, a biomedical data scientist at Stanford, is the corresponding author."- Hanae Armitage, SCOPE Blog (Stanford School of Medicine)


BMIR and DBDS Faculty Featured in New PanCan Papers

The Cancer Genome Atlas (TCGA) PanCan effort recently released a series of 27 papers. Published in various Cell journals, these papers mark the culmination and completion of the PanCancer Atlas Initiative (PanCan) and The Cancer Genome Atlas (TCGA) consortium. Faculty from Stanford’s Center for Biomedical Informatics Research (BMIR) and the Department of Biomedical Data Science (DBDS) contributed greatly to the completion of this large-scale PanCan initiative, which spanned a decade and analyzed over 11,000 tumors from 33 of the world’s most prevalent cancers.


Zou Group Receives a Google Faculty Research Award!

February 22, 2018- DBDS is proud to announce that Google will support the Zou group's research in developing new machine learning algorithms for genomics and health. The award will also enable collaborations with teams within Google. 


Germline determinants of the somatic mutation landscape in 2,642 cancer genomes

November 16, 2017- Francisco de la Vega (Adjunct Professor in DBDS) and (DBDS Chair) Carlos D. Bustamante have, for the past few years, been working on a project with the International Cancer Genome Consortium - the Pan-Cancer Analysis of Whole-Genomes (PCAWG) group. The project involves the analysis of whole-genome sequencing data from cancer tumor specimens from over 2,600 cancer patients and matched normal tissue, and hundreds of researchers around the world organized in subgroups over 14 research themes. Drs De La Vega and Bustamante have, more specifically, been participating in the analysis of germline genomes to understand how the germline variants affect, among other things, the cancer somatic mutational process. 

A milestone has recently been reached in the project, with the release of the germline team's manuscript at BioRxiv. This manuscript will be submitted soon for review, and will be published in a special issue in 3-4 months, with other consortium papers.


Interpretation of Neural Networks is Fragile

Fake news for AI by AI, in other words.

November 8, 2017- DBDS's Zou group just released a paper which, for the first time, demonstrates that interpretation of machine learning predictions are extremely fragile. The team, led by Ph.D. students Amirata Ghorbani and Abubakar Abid, showed that for two images that are visually identical and that are both predicted to be, for example, malignant, the machine learning algorithm can give two completely different explanations. 

This is very disconcerting in practice because in order to trust machine learning predictions, researchers typically rely on its explanation for why certain predictions are made. This work shows that the explanation itself is highly unreliable. 


Mosaic Mutations in Blood DNA Sequence Are Associated with Solid Tumor Cancers

July 6, 2017- This is the first of a couple of articles that will be coming out of the Rivas Lab and appearing in npj Genomic Medicine this summer 2017.

The editors of Genomic Medicine summarize the article as follows: "Having some abnormal blood cells with mutations that shorten the coding sequence of their genes increases one’s risk for solid tumors. Mark Daly and colleagues from the Broad Institute in Cambridge, Massachusetts, USA, used large genomic databases to test whether having blood cells both with and without genetic variants predicted to shorten the encoded protein — a phenomenon known as mosaic protein-truncating variants (PTVs) — was associated with developing a range of solid-tumor cancers. They studied DNA from around 8,000 people with cancer and 6,000 healthy controls. They confirmed previous reports linking these variants to breast and ovarian cancer, and extended the association to include tumors of the brain, skin and lungs. (Other studies have also shown that mosaic PTVs precede and predict the development of leukemia.) These results broadly connect cancer to blood DNA changes."


DBDS Chair and Faculty receive Chan Zuckerberg Investigator awards

February 8, 2017- DBDS Chair Dr. Carlos Bustamante and new DBDS faculty member James Zou are now recipients of the Chan Zuckerberg Biohub Investigator Award!

"The Chan Zuckerberg Biohub committed more than $50 million to support 47 of the best investigators from Bay Area universities, including 19 from Stanford University. The investigators each receive five-year appointments worth up to $1.5 million to carry out non-conventional scientific exploration and to invent new tools to accelerate the pace of discovery.

'The 47 CZ Biohub Investigators we’re introducing today are quite literally inventing the future of life science research,” said Stephen Quake, co-president of CZ Biohub and professor of bioengineering and applied physics at Stanford. “The CZ Biohub is distinguished by our emphasis on technology and engineering, and our researchers are inventing tools to accelerate science for the good of humanity.'"


Stanford Daily Interview with DBDS Faculty James Zou

February 3, 2017- "In this series, The Stanford Daily sits down to talk with new faculty members on campus.

Assistant professor James Zou joined Stanford’s newly created biomedical data science department in the fall, teaching CS 273B: “Deep Learning in Genomics and Biomedicine” with Anshul Kundaje, assistant professor of genetics and computer science. The Daily sat down with Zou to discuss joining the Stanford community and his interests beyond his academic discipline."


 

Carlos Bustamante Named Inaugural Chair of Newly Created Biomedical Data Science Department

September 11, 2015- "Population geneticist Carlos Bustamante will lead a new biomedical data department founded to advance precision health.

The department will build on the School of Medicine’s strengths in using information technology to advance precision health and work to find common solutions to the challenges of analyzing biomedical data from varied sources, including biosensors, electronic medical records and genomic sequencing."


Departments of Emergency Medicine, Biomedical Data Science To Be Created

June 11, 2015- "Biology and health care are being transformed by large-scale data analysis. The Department of Biomedical Data Science will build on the medical school’s strengths in developing and applying information technology to prevent disease, deliver more efficient patient care, streamline applications in translational research and improve access to biomedical data. The field has grown rapidly and has proven to be an invaluable tool for ensuring clinical objectives and best practices.

'Stanford is already a world leader in innovation methodology in biostatistics and biomedical informatics,' said Russ Altman, MD, PhD, professor of bioengineering, of genetics and of biomedical informatics research. 'Uniting these disciplines under the auspices of a single department will allow us to approach data and information in a whole new way.' Altman, a member of the core planning group that prepared the proposal for departmental status, added, 'Big data holds unbelievable opportunities to change how medicine is explored and delivered.'"