Long ago (see Early Life), I got my Bachelor of Arts in speech pathology and audiology and my first graduate degree in applied linguistics, both from Northern Arizona University. Then I became a chaplain (I'm an atheist and culturally Jewish by choice). I provided emotional and psychosocial palliative care for patients and families at New York-Presbyterian Hospital. I was 24 and struck by the inevitability of death and moved by the wails of the grieving and stories of the bedridden. I found myself wanting to be more than a listening ear, to prevent people from ever being in the hospital. Doing that, I reasoned, would’ve made me the best "chaplain." It would’ve been “divine” – an invisible intervention, a non-miraculous "miracle." But I didn’t know what that intervention was. I knew being a physician wasn't the answer because physicians treat existing illnesses. But I wanted to know what causes cancer, congenital disorders, psychiatric suffering, and diabetes, and to prevent them all.


Four years later, I learned there was an entire scientific discipline dedicated to fulfilling that wish: public health, a humanistic "miracle" that saves lives millions at a time. But life progresses moment by moment, and providing palliative care was my immediate focus back then. To be better at helping others, I strengthened certain psychological traits in myself through meditation. I took up residence as a monk at Green Gulch Farm, part of San Francisco Zen Center. For three years, I honed the ability to watch my thoughts, emotions, and sensory experiences. Though one of the tenets of Zen is that we aren't free, the goal, paradoxically, was to gain greater freedom, get some relief from the suffering of being run around by relentless thoughts. I experienced some of the peace. And I appreciated the deliberate life, the Japanese architecture, and the elegant, black robes. Yet after three years of sitting for many hours on floor cushions (like Leonard Cohen), I grew intolerably anxious. The longer I stared into the space of my mind, the more anxious I got! Although my chaplaincy training was rooted in Tavistock group dynamics (depth psychology) and my goal in practicing Zen was to learn about my mind to help others, I knew I didn't want to be a therapist. As Donald Winnicott, the English psychoanalyst wrote, "It is a joy to be hidden [like peace from Zen], but a disaster not to be found [wasting my intellectual life at a Zen center]." I knew I needed a doctorate.


So I left the cushion for a position that lit my imagination in Seattle. It was at a genetics laboratory that did population screening of newborns. Every baby born in Washington State got a “heel prick,” and their blood was screened for innate metabolic disorders. The aim was to identify and put sick babies on treatment before anyone noticed they were sick. Amazingly, I had stumbled into public health, an invisible way to intervene, to prevent babies, at least, from being hospitalized. I found an empirical way to be a chaplain that didn’t require being a chaplain – a way to invisibly reduce suffering.


While with the newborn screening program, I also sampled medical genetics and epidemiology, the study of diseases and their prevention in human populations. But getting just a taste of these wasn’t enough. I wanted formal training and left Seattle for a master’s in public health (MPH) at Johns Hopkins Bloomberg School of Public Health. And the fascination grew. Afterwards, I did a yearlong fellowship in the Clinical Genetics Branch of the National Cancer Institute, where I began studying rare cancers (testicular germ-cell tumors) with Dr. Mark Greene and cancer-predisposition syndromes (e.g., ribosomopathies) with Dr. Blanche Alter. Then I got my PhD in public-health genetics at the University of Washington.


My dissertation had two parts: one empirical and one in the humanities. I did the empirical half at the Fred Hutchinson Cancer Research Center in the epigenetics of shift work under Dr. Parveen Bhatti, an environmental epidemiologist. I did the humanities half on the ethical landscape of working at night with Dr. Wylie Burke, former head of American Society of Human Genetics. The dissertation products were published in three different scholarly venues: an epigenetics journal, an ethics journal, and Nature.


Several months before graduating, I moved to England to learn Mendelian randomization, a cutting-edge, causal-inference technique that uses genetics to study the environment, everything that isn’t genetic. Mendelian randomization mimics a randomized-controlled trial and avoids most sources of environmental confounding in observational studies.


I trained for two years at the University of Bristol, which specializes in Mendelian randomization, and published Mendelian randomization studies of metabolites and prostate cancer (here and here) before returning to the U.S. to a cancer hospital, City of Hope in Los Angeles. At City of Hope, I published more Mendelian randomization studies (e.g., on ovarian cancer, chronotype, Alzheimer's disease, and more) and began working with RNA-seq (gene expression) data. I'm now finishing my last years as a research fellow at the Harvard School of Public Health, where I'm weaving together Mendelian randomization, gene expression (transcriptomics), and nucleolar biology (ribosomal biogenesis).


Current Research


At Harvard in Bernardo Lemos’ lab, I have a number of projects underway involving ribosomal biology and longevity (see my slides for the development of ribosomal expression clocks of aging). I’ll tell you about one: autism spectrum disorder (ASD) as a ribosomopathy.




Our ribosomes are essential for life. They translate messenger RNA into proteins. This translational process is ultraconserved and heavily monitored by cells. Mutations in ribosomal genes can be fatal or lead to developmental disorders with craniofacial and limb abnormalities and cancer predisposition, a group of diseases known as ribosomopathies


ASD is also a group of developmental disorders that are heritable, though all the causes of ASD are not yet known. ASD is typically diagnosed within the first 1-3 years of life and is characterized by varying degrees of impairment in communication with or without intellectual disability. Like canonical ribosomopathies, those with ASD can have an array of craniofacial and other physical differences, though, with ASD, these are not typically as pronounced as with known ribosomopathies. Are some forms of ASD due to abnormal ribosomal biology?


Across 13 brain regions in a human cohort (18,381 cases of ASD & 27,969 controls), I've characterized the genetic control of expression of ribosomal proteins in relation to ASD. Cases of ASD were diagnosed in 2013 or earlier with one of five ICD10 codes: childhood autism (ICD10 code F84.0), atypical autism (F84.1), Asperger’s syndrome (F84.5), other pervasive developmental disorders (F84.8), and pervasive developmental disorder, unspecified (F84.9). Controls were children without ASD as of 2013.


But how did I get brain tissue from living people? I didn't! I performed transcriptomic Mendelian randomization, which enabled me to use the genetic data in brain tissues (from cadaver donors) and integrate it with the genetic data in the ASD cohort. Specifically, I obtained the brain data from the Gene-Tissue Expression (GTEx) project, which has done genetic studies of expression in tissues across the body. I extracted the genetic variants controlling ribosomal protein expression from the GTEx brain tissues. Then I extracted the same genetic variants in the ASD genetic study. I combined the genetic data for each variant to obtain the Mendelian randomization estimates. These estimates are considered "causal" if there is no unwanted pleiotropy or genetic confounding and if the instrumental genetic variant is strongly associated with gene expression. While a randomized-controlled trial is the gold standard for inferring causality in humans, Mendelian randomization avoids most sources of confounding in bioinformatics analyses and can provide candidate genes and mechanisms (i.e., expression) to study ASD experimentally in animal models. 


The forest plot below shows Mendelian randomization results (beta estimates) for one ribosomal protein (Gene 1, name withheld until published). The results are interpreted relative to an increase in expression, and the dotted line at zero represents the null of no association. None of the confidence intervals cross zero, meaning the results are significant at P<0.05. An increase in expression of the ribosomal protein increases risk for ASD, a finding that replicates across all 13 brain regions.


Ribosomal Gene 1 on ASD



Because the signal replicates across tissues, this strengthens the evidence that the above ribosomal protein is involved in the etiology of some forms of ASD. However, the confidence intervals are wide, perhaps because ASD is not a single disorder but a group of them with highly heterogeneous clinical presentation, including individuals with severe impairment and intellectual disability (ID) and those with above-average intelligence quotient (IQ) and high levels of educational attainment. The sample size for the ASD genetic study is also limited (n = 46,350). To address this, I've obtained the genetic data for a trait that is related to ASD but for which the sample size is much larger: education years (n = 293,723). Education years in this cohort was captured by mapping a years-of-education equivalent to the 1997 International Standard Classification of Education (ISCED) codes for each participant who was at least 30 years old. That is, education years is a standardized measure of the number of schooling years a participant had up until age 30 (inclusive of primary school and college). 


ASD and education years are positively genetically correlated (genetic correlation = 0.19). This means that some with ASD, likely those with Asperger's syndrome (not those with intellectual disability) are predisposed to stay in higher education longer. In the (less colorful) forest plot directly below for Ribosomal Gene 1 are the Mendelian randomization results for the same gene as above (whose name I'm still concealing) in relation to both ASD and education years. The results for ASD, which you have already seen, are in black (again notice the wide confidence intervals), and the results for education years are in red (notice the tight confidence intervals). The comparison I want you to see is that the results are null for education years. Same as above, the confidence intervals for ASD do not cross zero; however, for education years, they do. This means, for this gene, that aspect of ASB influenced by ribosomal protein expression may be developmental but not cognitive, since the genetic variant doesn't impair or improve education years, which is highly correlated with cognitive ability.


Ribosomal Gene 1 on ASD (black) and education years (red)



The story is different for Gene 2 below. For this ribosomal protein, we still observe wide confidence intervals for ASD (in black), and they cross zero. But for education years (in red), the confidence intervals are tight, and the estimates are multiple-testing significant. An increase in the expression of this ribosomal protein is neuroprotective, increasing education years. Here, although the estimates for ASD cross zero, the magnitude of their effects matches those for education years, possibly indicating that the signal for ASD being captured here is for Asperger's syndrome (without cognitive impairment and/or higher cognitive ability). 


Ribosomal Gene 2: On ASD (black) and education years (red)