Our travels to the Big Island to discuss ethical and technical challenges in genome privacy
by Charlotte Brannon
Gamze and I started off 2020 with a bang by attending the Pacific Symposium on Biocomputing (PSB) on the Big Island of Hawaii. This year was the 25th anniversary of PSB. The conference was held on the west coast of the island at the Fairmont Orchid hotel. We attended several of the conference sessions and workshops, but we were most excited about the “Navigating Ethical Quandaries with the Privacy Dilemma of Biomedical Datasets” workshop because Gamze was a co-organizer and I was giving a talk. I also took a bunch of photos, which are not *technically* related to ethics, privacy, or genomics, but nonetheless accompany this post.
left- arriving at the Kona airport middle - views of the island on the drive to my hotel right - arriving at the conference on day 1
“Ethical quandaries” come up a lot in computational biology and bioinformatics in more ways than one. In fact, ours was not the only workshop at PSB this year that addressed questions of ethics. One of the other workshops focused on Artificial Intelligence (AI) ethics in Biomedicine. Whereas Gamze and I are specifically interested in protecting individual privacy while sharing biomedical datasets, AI presents a different set of ethical challenges. For example, scientists making use of AI technology have to think about how to prevent bias in their algorithms. This is a tricky question, as AI/machine learning algorithms make use of real-world data to learn how to approach new data. Yet, real-world data is often biased, which will lead to a biased algorithm. Humans are inherently biased, but at least we can hold people socially and legally accountable for acting on biases. If someone were to develop a machine learning algorithm to, say, identify cancer in patient imaging data, and it were biased toward certain populations, we currently wouldn’t have a good way of holding that algorithm “accountable” for its bias.
Additionally, algorithms that can identify hidden patterns in large amounts of data don’t always get it right. In the AI ethics workshop, Chris Re from Stanford gave an example of a machine learning algorithm that was supposed to determine a person’s sex based on an image of their eye. At first, the algorithm seemed to be incredibly accurate. It turned out, though, that the algorithm was determining sex based on the presence or absence of mascara on the eyelashes, which were visible in each photo in the training data set. Imagine if we had algorithms like this making decisions about health care diagnoses and treatments… For this reason, it was important to have a workshop to discuss ways scientists can continue to refine and make use of AI technology while addressing ethical issues. This workshop reminded us of the amazing invited talk given by Julia Stoyanovich at the 2019 GA4GH plenary meeting, who advised the audience to “data responsibly,” as the world is biased.
Here’s another “ethical quandary” – how can we allow researchers to share biomedical data and make groundbreaking scientific discoveries without violating individuals’ privacy? This was the subject of the “Navigating Ethical Quandaries with the Privacy Dilemma of Biomedical Datasets” workshop, which was co-organized by both technical experts (like Gamze, a postdoc in computational biology and bioinformatics focusing on genome privacy and Dr. Steven Brenner, who is a Professor of Computational Biology), and legal or ethical experts (like Jennifer Wagner, a JD/PhD focusing on anthropology, genetics, law, and bioethics; Megan Doerr, a governance specialist at the Sage Bionetworks, and John Wilbanks, who is the the chief commons officer at Sage Bionetworks). At PSB 2019, Gamze and Steven (together with other technical experts) held a session focused on privacy, which was very successful and ended with a fruitful panel discussion. Based on how well it went, they submitted a workshop proposal for 2020 focused mainly on privacy. Simultaneously, Megan Doerr (from Sage Bionetworks) and others submitted a proposal for an ethics workshop, and the conference organizers asked them to merge into one workshop.
Megan opened this year’s workshop by saying that the marriage of technical and ethical minds in this field used to be a “marriage of convenience,” but now it is a “marriage of love.” This got a big laugh, but we actually ended up circling back to this analogy throughout the workshop. What does a “marriage of love” mean between ethics and technical tools when sharing biomedical datasets? To me, it could mean a few things, but at the very least it meant that as technical and legal/ethical experts, we were actually going to engage with one another, rather than simply talk at each other. Privacy and ethics cannot be separated in these discussions.
The workshop consisted of several talks, some presenting technical tools/algorithms for protecting patient privacy when sharing biomedical data, and others discussing the ethics of biomedical research. For example, Corey Hudson from Sandia National Labs gave a talk about hackable vulnerabilities in genomic analysis pipelines; I gave a talk about applications of blockchain technology to promote security of genomics data; and Jennifer Wagner gave a talk about legal regulation of diverse biomedical datasets. Lucila Ohno-Machado, Heidi Sofia and Xiaoqian Jiang mentioned the iDASH center and challenges (which we discussed in our previous blog post). The complete list of talks and speakers is as follows:
John Wilbanks, Translational bioethics in a monopoly network era
Jennifer Wagner, Legal Angles and the ‘Illusion’ of Certainty: Regulation of Diverse Data Sets
Heidi Sofia, Advancing data sharing with security and privacy at NIH
Lucila Ohno-Machado, Responsible data sharing: Patient preferences, institutional policies, and privacy technology
Corey Hudson, From buffer overflowing genomics tools to securing genomic pipelines
Charlotte Brannon, Applications of blockchain technology to genomics
Xiaoqian Jiang, Secure Cohort Identification for Clinical Trial using Heterogeneous Healthcare Data
After this mix of technical and legal/ethical talks, we had a panel discussion with all of the speakers, which brought us back to some foundational issues at the intersection of these two modes of thought (the technical and ethical). One question that came up was, what does it mean to share data legally versus ethically? What should the law do to protect patient privacy while permitting biomedical research to move ahead? Megan Doerr brought up the issue of informed consent. As privacy issues remain, we want people to be informed before sharing their data. Yet, achieving truly informed consent is challenging. For example, Megan pointed out, if you ask people whether they are willing to share their “genomic data,” this may not meet the standard for informed consent–many people do not know what the term “genomic data” means. However, it turns out that most people do know what “DNA” means, and therefore it may be a more appropriate term to use when obtaining consent for data sharing. This is a case where asking about “genomic data” might be checking the box of getting consent, but we can do so more ethically by using terms that people really understand. There were also discussions about the problems with broad consent vs. specific consent.
As a quick tangent, I even saw examples of this during my time on the Big Island. Over the course of the conference, I had to take several taxi rides between the conference and my hotel a little ways down the coast. Most of the drivers asked what conference I was attending, and what kind of research I was involved in – they were used to meeting people who had come to the Big Island for a conference. In these conversations, it was challenging to find the appropriate jargon and approach to discuss technical topics. Most people I chatted with did not know what I meant by “genome privacy” (the term Gamze and I use colloquially in the lab), but most people understood when I said something like, “looking for ways to share people’s DNA data for scientific research without violating personal privacy.” Actually, most people were eager to discuss this with me for the rest of the car ride! One woman and I actually had a 30 minute conversation about how machine learning works, and what sorts of ethical problems the field of AI faces. This showed me that people are interested in these topics, if they are made accessible.
One guy I met was especially interested in the focus of our workshop. Upon hearing that I was attending a workshop about DNA privacy, he immediately told me that in Hawaii, people are especially invested in this issue–that Native Hawaiians in particular are weary of using DNA to determine ancestry. I had never heard this before, but later found this 2015 article, which reports a controversy over using blood quantum to determine Hawaiian identity. The article quotes Williamson Chang, a Native Hawaiian law professor at the University of Hawaii, who says it is a very “un-Hawaiian idea” to rely on DNA to define identity; that personal identity and family heritage are more important than the percent composition of blood.
I found this interesting because it shows that different populations have different stakes when it comes to genome privacy – we need to be able consider this diversity when making privacy laws and standards, or designing processes like informed consent. I appreciated the chances I had at the conference to talk to several non-scientists about the implications of genomic privacy protection.
At one of the conference social events after the workshop, a fellow trainee asked me how I enjoyed the workshop, did I think it was productive, and if I could wave a magic wand and achieve three things in the field, what would I do? I almost wish this had been the core question of the workshop panel. What three things could we, a mix of lawyers, scientists, and ethicists agree on that would move the field forward? I couldn’t answer the question by myself, and I’m not sure we would have identified three things within the 45 minutes for our panel. We have many lingering questions, which we hope to come back to next year:
- There are weak or no consequences for privacy violators. How can we impose sanctions so that people and organizations will follow privacy law?
- What is currently going on in data sharing that is legal, but not ethical? We want to hear from more speakers who can address this.
- How can we do a better job of moving privacy-protecting tools into real-world scenarios?
Next year at PSB, we hope to have a chance to discuss some of these questions, and more. In the meantime, we wanted to acknowledge this year’s workshop organizers, listed below.
Workshop Organizers: Gamze Gursoy, Megan Doerr, Steven E. Brenner, Haixu Tang, John Wilbanks, Jennifer K. Wagner