Skip to content Skip to navigation

Our Science Distilled

Plain language summaries of our scientific publications

Mobile health monitoring of cardiac status

Cardiovascular diseases (CVD) are diseases of the heart and blood vessels which provide the body with blood and oxygen to support life. More people are sick and die from cardiovascular diseases than any other diseases. A myocardial infarction, or heart attack, is often caused when a person with coronary artery disease, or a narrowing of the blood vessels which supply the heart with blood and oxygen. A stroke is when the blood vessels supplying the brain with blood and oxygen become clogged or damaged. Monitoring patients with CVD can save lives by preventing the diseases from progressing or by responding quickly to events when they occur.

Smart technology and mobile health platforms are creating an environment where timely prevention of and response to cardiac events is becoming a reality. For example, wearable monitors range from smart watches which can detect dangerous irregular heart rhythms to implantable defibrillators which can save the life of someone having a heart attack with a therapeutic shock and automatically inform a patient’s doctor. Other monitors can measure physical activity, sweat production, swelling, blood pressure, stroke volume and pulse wave velocity, which are all important in detecting heart disease. In the future, these may be integrated into a “smart medical home” which could provide automated notification to a healthcare team, including the exact location of the patient in case of a medical emergency.

Christle, Hershman, Soto, Ashley (2020) Annual Review of Biomedical Data

View Article

Stretch-Induced Biased Signaling in Angiotensin II Type 1 and Apelin Receptors for the Mediation of Cardiac Contractility and Hypertrophy

The heart is able to sense the physical stress from blood flow and adapt to the body’s needs. Too much stress, however, can cause the heart not to adapt properly, leading to heart failure and abnormal heart rhythms. Physical stress causes the heart to send signals that are both helpful and harmful. Because of this, researchers have been trying to find drugs that stop the harmful signals, leaving only the helpful signals.

In this review, we discuss how the heart reacts to physical stress in ways that are good, and in ways that are damaging. We also talk about how it might be possible to study the helpful and harmful signals sent by the heart to find new drugs for heart failure.

Seo, Parikh, Ashley. (2020) Frontiers in Physiology

View Article

High-throughput SARS-CoV-2 and host genome sequencing from single nasopharyngeal swabs

During a disease outbreak like the COVID-19 pandemic, genetic information about both the new virus and the people who become infected can be very valuable. Learning what is written in the genome of the virus (encoded in the molecule RNA) can help us track its spread among populations and understand how it is evolving. Studying the genome of humans who are infected (encoded in a similar molecule, DNA) can help us better understand who is more at risk for severe disease. However, collecting all of this data while also managing the massive public health demands of a pandemic is a major logistical challenge. In particular, it requires a lot of resources to conduct separate research studies on COVID-19 while the main priority is diagnosing and treating patients. To help solve this problem, we came up with a strategy to collect information on the virus RNA and human DNA from the nasal swab that is left over after a positive COVID test. We dissolve the material still on the swab into a liquid, and then divide it up for extractions of RNA and DNA. We then read the genetic code of this extracted RNA and DNA using multiple different technologies, collecting several types of data about both the virus and the human host. This approach will allow us to gather important data during the pandemic without collecting additional samples just for research, and it can also be used to study disease outbreaks in the future.

Gorzynski, De Jong et al. (2020) MedRxiv

View Article

Targeted Long-Read RNA Sequencing Demonstrates Transcriptional Diversity Driven by Splice-Site Variation in MYBPC3

Finding DNA variants that cause disease can be important to help diagnose and treat patients. But sometimes, patients have new variants that we haven’t seen before. One way that we can deduce if the variant causes disease is to look at how it affects RNA. When DNA is transcribed into RNA, sections called introns are spliced out to leave just the parts that will be turned into protein, called exons, behind. Variation in DNA can disrupt this splicing process, leading to RNA that will create a dysfunctional or nonfunctional protein, which could cause disease. In patients with rare variants, we can use RNA sequencing to look at whether or not these variants affect splicing.

We developed methods to examine RNA splicing in clinical cardiac samples. While this has been done before, it usually only looks at short, incomplete fragments of the RNA. We used long read sequencing to look at the full length of an RNA transcript all at once. We tested this method on samples from a patient with hypertrophic cardiomyopathy. We found that the novel mutation in our patient created lots of dysfunctional RNA. In fact, for this particular mutation we expected to find only a single alternative version of the RNA transcript, but instead identified at least 11. This gave us strong evidence that this mutation could be causing the disease. Implementing these methods in patients with rare variants can provide targets for potential therapies as well as aid in screening to find other potentially affected family members and keep them safe.

Dainis et al. (2019) Circulation: Genomic and Precision Medicine

View Article

Multi-task deep learning for cardiac rhythm detection in wearable devices

The rise of fitness devices like the Apple watch, Fitbit, Oura ring and many others aim to keep us healthy. Most of these devices measure heart rate and step counts while, increasingly, some go further to try to detect disease. Our paper explores how we can combine wearable devices with deep learning “artificial intelligence” methods for the real time detection of a common abnormal heart rhythm known as atrial fibrillation.

Atrial fibrillation (AF) affects upwards of one in five people in their lifetime. It describes chaotic electrical activity of the top chambers of the heart, the atria. It can cause palpitations but, if it goes unnoticed as it commonly does, it can be especially dangerous through its risk of stroke. Detecting AF early and preventing stroke could save millions of dollars and prevent major disparities in care.

Wearable devices are ideally suited to detect atrial fibrillation because they measure heart rate by intermittently shining light into the skin. A sensor in the wearable then detects a signal from returning light which changes in its intensity as the blood from each beat passess beneath it.

In this work, we collected over one million computer-generated signals and gathered datasets of real signals from over one hundred individuals from three different wearable devices to design an algorithm, DeepBeat.

We found the following: [1] Incorporating information from the signal quality, i.e. how clear the signal is, allows the algorithm to tell apart a true case of AF from a non-case. In comparison, without considering the quality of signal, a false positive may occur due to noisy/unclear signals. [2] We used a large amount (>1 million) of computer-generated signals to pre-train the model. Pre-training allows the algorithm to learn important aspects of the signals from the computer-generated signals before it ever starts looking at real signals. Once real signals are introduced to the algorithm, the algorithm learns much faster and better, and our results improve substantially.

Lastly, we wanted to test our algorithm, DeepBeat, on a set of individuals completely different from those we used to train our model. We found that our model correctly generates a positive result for people who have AF 98% of the time. Our algorithm serves as a foundational step towards early detection of AF and a reduction in the financial and emotional trauma of stroke.

Torres Soto, Ashley (2020) npj Digital Medicine

View Article