The Bioinformatics CRO Podcast

Episode 51 with Adam Freund

Adam Freund, founder and CEO of Arda Therapeutics, discusses how targeted killing of pathogenic cells could be used to treat chronic disease and aging. 

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Adam is founder of and CEO of Arda Therapeutics, a company using single-cell sequencing to characterize and target pathogenic cells to treat chronic diseases and aging.

Transcript of Episode 51: Adam Freund

Coming soon…

States Ranked by Age adjusted COVID Deaths - Updated April 20 - See table for more details

States Ranked by Age-Adjusted COVID Deaths

Data updated on December 7, 2022

We’re inundated with statistics on how US states have fared relative to one another throughout the pandemic. Sometimes these can appear contradictory because the data can be cut to support a variety of narratives. We wanted an updated source of cumulative age-adjusted COVID-19 deaths by state, as death is an important measure of the impact of a pandemic, and states have adopted widely divergent policies.

While COVID-19 deaths are usually adjusted for state population (deaths per 100,000), they are usually not adjusted for the age distribution of a state. It’s important to adjust for age when considering state-to-state differences in outcomes as age is the dominant risk factor for death provided someone is infected with COVID, and state age distributions vary considerably. In the following analysis, we present age-adjusted cumulative COVID-19 deaths and rank each state plus Puerto Rico and the District of Columbia accordingly. We generated the plot and table using the CDC’s Provisional COVID-19 Death Counts by Sex, Age, and State database, which sourced its data from death certificates. These numbers are more consistently processed across states, though they may differ slightly from other sources. 

 

Bubble areas are proportional to state populations and the horizontal arrangement is arbitrary to reduce overlap.

Here we see dramatic state-to-state differences in cumulative age-adjusted COVID deaths per capita to date, spanning a range of over five fold. In the end, some states that adopted dramatically divergent policies had comparable outcomes (Florida and California, for example).

Mississippi is exceptionally high. A few regional clusters have fared markedly better than the rest: Vermont, New Hampshire & Maine, Oregon & Washington, and Hawaii & Puerto Rico.

Why do COVID deaths vary by state? 

Explore the relationships between age-adjusted COVID deaths and several state-level metrics including: vaccination coverage, obesity rate, the strictness of COVID policy and more. 

Explore the Data

 

Age-Adjusted COVID Deaths Ranking

State

COVID-19 Deaths per 100,000

Age-Adjusted COVID-19 Deaths per 100,000

1 Mississippi 474 486
2 Oklahoma 438 447
3 Kentucky 416 420
3 Tennessee 415 420
5 Texas 344 412
6 Alabama 417 410
7 Nevada 379 403
8 Arkansas 397 385
8 New Mexico 399 385
10 Indiana 377 384
11 Ohio 399 380
12 Louisiana 363 378
13 North Dakota 388 377
13 West Virginia 430 377
15 Arizona 394 375
16 Georgia 325 371
17 New York 391 370
18 South Carolina 378 369
19 New Jersey 381 367
20 District of Columbia 300 354
21 South Dakota 371 351
22 Missouri 359 343
23 Pennsylvania 389 342
24 Michigan 349 333
25 Montana 353 324
25 Rhode Island 363 324
27 Kansas 327 321
28 Idaho 299 315
29 North Carolina 303 308
30 Wyoming 301 302
31 Florida 357 299
32 Iowa 327 298
33 Delaware 326 297
34 Connecticut 330 295
35 Illinois 289 289
36 Maryland 278 286
37 Colorado 250 285
38 California 258 275
39 Nebraska 276 272
40 Massachusetts 287 270
41 Virginia 251 262
42 Wisconsin 273 260
43 Alaska 192 253
44 Minnesota 250 246
45 Utah 173 231
46 Oregon 197 190
46 Washington 179 190
48 New Hampshire 205 189
49 Maine 215 180
50 Puerto Rico 167 143
51 Vermont 130 114
52 Hawaii 120 106

Calculation 

We determined the age adjusted mortality per 100,000 people (maa) for each state using the formula:

m_aa = SUM (D_x * P_x / (N_x * 100,000))

Where Dx is the total deaths in age group x in the state, Nx is the total population in age group x in the state, and Px is the percent of the population in age group x in the United States.

Citations

COVID deaths from CDC: “Provisional COVID-19 Death Counts by Sex, Age, and State” (Updated on December 7, 2022)

Population data from U.S. Census Bureau: “State Population by Characteristics: 2010-2019”

The Bioinformatics CRO Podcast

Episode 50 with Alfredo Andere

Alfredo Andere, co-founder and CEO of Latch Bio, discusses the unique challenges facing young entrepreneurs and the future of cloud computing in biology. 

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Alfredo is co-founder of and CEO of Latch Bio, a cloud bioinformatics platform that enables collaboration between computational biologists and wet lab researchers.

Transcript of Episode 50: Alfredo Andere

Coming soon…

The Bioinformatics CRO Podcast

Episode 49 with Joshua Hare

Joshua Hare, Professor of Medicine at the University of Miami and co-founder of Longeveron, discusses the regenerative and reparative potential of MSCs and how cell therapies will revolutionize medicine.

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Joshua is professor of Medicine at the University of Miami and co-founder of Longeveron, a biotech company using MSCs to treat chronic diseases. He is also founding director of the Interdisciplinary Stem Cell Institute at the University of Miami’s Miller School of Medicine. 

Transcript of Episode 49: Joshua Hare

Coming soon…

Why do COVID Deaths Vary by State?

Cumulative COVID deaths are an endpoint to compare the effectiveness of states’ pandemic-related health policies. As the major risk factor for death from COVID is age, and some states have younger populations than others, it’s important to adjust for age when trying to understand factors that contribute to variable outcomes among states. Here we examine the relationships among age-adjusted COVID deaths and several variables of interest. Note that such comparisons cannot, on their own, be used to identify causal contributors to cumulative age-adjusted COVID deaths. Many features are highly co-linear and may serve as an imperfect proxy for underlying causes. However, these plots can be useful for hypothesis generation, and can support (but not prove) the lack of a strong causal relationship when there appears to be no association at all. We recommend bearing in mind the ecological fallacy.

Unadjusted COVID deaths versus Age-adjusted deaths | y = 0.934 * x + 15.3 | R^2 = 0.912

The chart above shows the relationship between age-adjusted COVID deaths and unadjusted deaths. States with younger populations such as Texas perform worse after age adjustment, while states with older populations such as Florida perform better.

Summary 

  • Vaccination rate is the single biggest predictor of age-adjusted deaths by state.
  • COVID deaths do not correlate with state stringency after adjusting for age and obesity rate.
  • State stringency correlates with unemployment rate but not with non-COVID excess deaths, depression or suicide vs baseline.
  • Partisanship is a strong predictor of vaccination rate, but not of deaths after accounting for age, vaccination rate and obesity.

This is incomplete and there are many factors as yet unexamined. We will add more visualizations over time.

 

Vaccination

Vaccination is highly effective in preventing severe outcomes and death from COVID infection. Indeed state vaccination rate has a strong negative correlation with age-adjusted COVID deaths (P<0.01).  In the 65+ age group, there is a similarly high correlation (P<0.01) with age adjusted COVID deaths. Note that the CDC caps vaccination coverage metrics at 95%, due to issues tracking first and second doses. Because vaccination over 65 is the strongest single predictor of age-adjusted deaths, we also present cumulative COVID deaths adjusting for both age and vaccination rate over 65.

Vaccination Rate and Age Adjusted COVID Deaths

Fully Vaccinated versus Age-adjusted deaths | y = -555 * x + 645 | R^2 = 0.4

Vaccination Rate Over 65 Years of Age

Fully vaccinated over 65 versus Age-adjusted deaths | y = -1113 * x + 1282 | R^2 = 0.39

Vaccination Rate 65+ and Age & Obesity Adjusted COVID Deaths

Vaccination over 65 versus Obesity and Age-adjusted deaths | y = -527 * x + 770 | R^2 = 0.144

There is a moderate negative correlation (P<0.01) between vaccination and COVID deaths when adjusting for age and obesity. The reduction in correlation is due to collinearity between state obesity rates and vaccination rates. That is, states with higher adult obesity rates tend to have lower vaccination rates (P<0.01). In a multivariate linear model of age-adjusted deaths, vaccination over 65, and obesity, both showed significant correlation with deaths and had a multiple R² of 0.48. 

Stringency Index

Stringency Index is a useful tool for comparing government response to the pandemic. It is the average of nine policy metrics, including: 

  • School closures
  • Workplace closures
  • Cancellation of public events
  • Restrictions on public gatherings
  • Closures of public transport
  • Stay-at-home requirements
  • Public information campaigns
  • Restrictions on internal movements
  • International travel controls

The higher the stringency index, which has a maximum value of 100, the stricter a government’s response to the pandemic. Stringency index does not factor in compliance with government policy. For more information on Stringency Index and its calculation visit ourworldindata.org/ 

The charts above show a moderate negative correlation between the average strictness of pandemic-related policy and age adjusted COVID deaths (P<0.01) and a weak negative correlation between vaccination and age adjusted COVID deaths (Vaccination 65+ adjusted P=0.045). However, when deaths are additionally adjusted for obesity, there is no longer a significant correlation (Obesity adjusted P=0.16, Obesity and Vaccination 65+ adjusted P=0.22). Although some studies have suggested policies are effective in reducing COVID deaths and reducing strain on hospitals (1, 2, 3), effectiveness may decline over time with reduced compliance as more people are experiencing “lockdown fatigue“. Additionally, a recent meta-analysis found effectively no impact of lockdowns on mortality. As we are visualizing cumulative deaths across the pandemic, we will not see short-term impacts.

Mean Stringency Index and Adult Depression & Anxiety 

Mean Stringency Index versus % Change in Anxiety/Depression | y = -0.0058 * x + 0.639 | R^2 = 0.04

Stringency Index and Suicide

Mean Stringency Index versus % Change in Suicide Mortality | y = 0.0017 * x + 0.013 | R^2 = 0.024

All states in the US observed an increase in depression and anxiety symptoms over the course of the pandemic. However, these increases were not significantly correlated with mean stringency index (P=0.16).  Further, mean stringency index was not correlated with the % change in suicide mortality between 2018 and 2021 (P=0.28). 

Stringency Index and Unemployment 

Mean Stringency Index versus Change in Seasonally adjusted % Unemployment | y = 8.66E-4 * x + 0.0185 | R^2 = 0.28

Stringency index was moderately correlated with the % change in unemployment from the start of the pandemic to Sept 2021 (P<0.01) For further reading about the economic impacts of the COVID pandemic and government interventions see the following articles: The COVID-19 crisis: what explains cross-country differences in the pandemic’s short-term economic impact?, Epidemiological and economic impact of COVID-19 in the US, and
Pandemic Impact on Mortality and Economy Varies Across Age Groups and Geographies.

Obesity

Our data show that adult obesity rate has a moderate positive correlation with age adjusted COVID deaths both before and after adjusting for vaccination status (P<0.01 and P<0.05 respectively). This is consistent with prior studies which identify obesity as a significant risk factor for death from COVID infection (OR = 1.61). However, the slope of these plots is a few fold higher than one would expect given the individual-level risks of obesity alone even after adjusting for age & vaccination, which suggests that state obesity rate may serve as a proxy for other population-level risk factors for COVID death.

Obesity and Age Adjusted COVID Deaths

Adult Obesity Rate versus Age-adjusted deaths | y = 1198 * x -97.8 | R^2 = 0.371

Obesity and Age & Vaccination 65+ Adjusted COVID Deaths

Adult Obesity in 2021 versus Age and Vaccination adjusted COVID deaths | y = 523 * x + 122 | R^2 = 0.117

Poverty & Income Inequality

The Gini Index represents the degree of inequality in the distribution of income in a particular location. It’s value ranges from 0 to 1 with higher values indicating greater income inequality. You can find a more detailed explanation of its calculation here. The charts below show a significant (P<0.01) positive correlation between income inequality and age-adjusted COVID deaths, even after adjusting for obesity and vaccination. 

We also examined the poverty rate, using the supplementary poverty measure (SPM). The SPM takes into account differences in regional cost of living as well as taxes and the value of government assistance programs. Click here for more information about its calculation and comparison to the official poverty measure. Although Gini Index and the SPM are highly correlated with one another, it appears that Gini Index demonstrates a higher correlation with COVID mortality, especially after adjusting for vaccination and obesity. 

Partisanship

We used the percent of the population who voted for Trump in the 2020 presidential election to examine associations with partisanship. Although this will not be directly causal of COVID deaths, there has been a partisan aspect to vaccination in the United States, which does influence risk of death from COVID infection. Our analysis shows a very strong negative correlation between vaccination status and percent Trump vote.  Additionally, although there was a strong correlation between obesity and Trump vote, in a multivariate linear model with obesity, and vaccination over 65, percent Trump vote was not significantly associated with age-adjusted COVID deaths (P=0.76).

% Trump Vote (2020) and Vaccination Rate

% Trump Vote versus Fully Vaccinated | y = -0.756 * x + 1.02 | R^2 = 0.775

% Trump Vote and Vaccination Rate 65+

% Trump Vote versus % Fully Vaccinated Over 65 | y = -0.282 * x + 1.03 | R^2 = 0.442

Most states showed high vaccination rates in people aged 65 and older, and significantly reduced vaccination rates in younger populations. Although states with high Trump vote had significantly lower vaccination rates in the 65+ age group (P<0.01), the difference was exaggerated in younger populations (P<0.01). This is also shown in the chart below (P<0.01). 

Difference Between % Fully Vaccinated Over 65 & Under 65 and Trump Vote

% Trump Vote versus % Vaccinated Over 65 Minus Under 65 | y = 0.519 * x - 0.057 | R^2 = 0.674

Stringency Index and Age Adjusted COVID Deaths (Colored by Governor Partisanship)

Summary Statistics for States with Democrat Governorss

Mean Age-adjusted COVID Deaths/100K = 272±77 (std dev)

Mean Stringency Index = 40.4±6.3 (std dev)

Summary Statistics for States with Republican Governors

Mean Age-adjusted COVID Deaths/100K = 305±79 (std dev)

Mean Stringency Index = 31.4±6.0 (std dev)

Race & Ethnicity

Members of some racial and ethnic minority groups are more likely to experience severe outcomes from COVID-19 infections. The prevalence of underlying comorbidities that increase risk of severe outcomes from COVID infection (i.e. cardiovascular disease, asthma, obesity) vary by race and ethnicity and they may face differences in access to adequate healthcare resources.  As the racial composition of Hawaii is significantly different from the rest of the United States (i.e. an outlier), it was excluded from this analysis. 

Percent Population White versus Age-adjusted deaths | y = -555 * x + 645 | R^2 = 0.4 It should be noted that the race and ethnicity data are not normally distributed (Shapiro Test P<0.01 for all). White P<0.01 | Black P<0.01 | Asian P=0.22 | Indigenous P=0.66 | Hispanic P=0.30
Age-adjusted deaths versus % Population Black | y = -2.7E-4 * x + | R^2 = 0.214

Race & Ethnicity and COVID Deaths Adjusted for Age, Vaccination Over 65 & Obesity

Percent Population White versus Age-adjusted deaths | y = -171 * x + 426 | R^2 = 0.114 It should be noted that the race and ethnicity data are not normally distributed (Shapiro Test P<0.01 for all). White P<0.05 | Black P=0.19 | Asian P=0.26 | Indigenous P=0.70 | Hispanic P<0.01
Age-adjusted deaths versus % Population Black | y = 105 * x + 282 | R^2 = 0.036
% Population Asian versus Age-adjusted deaths | y = 316 * x + 282 | R^2 = 0.027
% Population Indigenous versus Age-adjusted deaths | y = 94.3 * x + 291 | R^2 = 0.003
% Population Hispanic versus Age-adjusted deaths | y = 238 * x + 263 | R^2 = 0.212

In a multiple regression model with obesity and vaccination over 65, % Hispanic population was significantly positively correlated with age adjusted deaths (P<0.05).

Mental Health

There are many new stresses associated with the pandemic that could lead to increases in mental illness in the population: worry about health, death of a loved one, chronic symptoms from long COVID, increased unemployment, and social isolation. All states observed an increase in depression and anxiety symptoms over the course of the pandemic and many had increases in suicides. However, based on our data these increases were not correlated with COVID deaths (depression & anxiety P=0.76, suicide P=0.12) or stringency index (depression & anxiety P=0.16, suicide P=0.28).

Age Adjusted COVID Deaths and % Change in Anxiety & Depressions Symptoms

Age-adjusted deaths versus % Change in Anxiety/Depression | y = -7.64E-5 * x + 0.454 | R^2 = 0.001

No Correlation P=0.76

Age Adjusted COVID Deaths and % Change in Suicide Mortality 

Age-adjusted deaths versus % Change in Suicide Mortality | y = -2.7E-4 * x + 0.126 | R^2 = 0.067

No Correlation P=0.12

% Change in Unemployment and % Change in Anxiety & Depressions Symptoms

% Change in Unemployment versus % Change in Anxiety/Depression | y = 0.935 * x + 12.9 | R^2 = 0.915

The chart to the left shows no significant correlation between anxiety and depression symptoms and the difference in seasonally adjusted unemployment from Jan 2020 to Sept 2021 (P=0.053)

Excess Deaths

Excess death is the difference between observed deaths in a specific time period and expected deaths in the same time period. The data below show the excess deaths in each state from Jan 2020 to Dec 2021. Expected deaths are based on historical trends from 6 years prior to the initial outbreak. The data below show a negative correlation between excess deaths and vaccination over 65 (P<0.01), which may be indicative that some deaths coded as non-COVID may in fact be related. For example, several studies have shown that even mild COVID infection significantly increases a person’s risk of cardiovascular disease. One found that the risk of stroke was 52% higher and heart failure was 72% higher in people who had been infected with COVID than those who had not. 

Excess Deaths and Vaccination Rate

Including COVID Deaths y = -0.252 * x + 0.291 R^2 = 0.218 | Excluding COVID Deaths y = -0.10 * x + 0.072 R^2 = 0.073

Including COVID Deaths:  P<0.01 | Excluding COVID Deaths: P=0.057

Excess Deaths and Vaccination Rate 65+

Including COVID Deaths y = -0.627 * x + 0.688 R^2 = 0.376 | Excluding COVID Deaths y = -0.278 * x + 0.255 R^2 = 0.156

Including COVID Deaths:  P<0.01 | Excluding COVID Deaths: P<0.01

Excess Deaths and Mean Stringency Index

Including COVID Deaths y = -0.0016 * x + 0.195 R^2 = 0.071 | Excluding COVID Deaths y = -2.2E-4 * x + 0.019 R^2 = 0.003

Including COVID Deaths:  P=0.061 | Excluding COVID Deaths: P=0.71

Excess Deaths and Obesity

Including COVID Deaths y = -0.296 * x + 0.042 R^2 = 0.066 | Excluding COVID Deaths y = 0.0681 * x - 0.0112 R^2 = 0.007

Including COVID Deaths:  P=0.071 Excluding COVID Deaths: P=0.56

Notes on Methods

  • P-values for plots were calculated using the Fisher R to Z transformation
  • P-values for multivariate models were calculated by multiple regression
  • Significance level was set at 0.05

Citations

The Bioinformatics CRO Podcast

Episode 48 with Alex Shalek

Alex Shalek, Associate Professor of Chemistry at MIT, discusses new methods and technologies in systems biology that have enabled advances in the diagnosis and treatment of autoimmune diseases and cancer. 

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Alex is Associate Professor of Chemistry at MIT, where his multi-disciplinary research aims to create and implement broadly-applicable methods to improve prognostics, diagnostics, and therapeutics for autoimmune diseases and cancer. 

Transcript of Episode 48: Alex Shalek

Coming soon…

The Bioinformatics CRO Podcast

Episode 47 with Jamie Smyth

Jamie Smyth, Associate Professor at Virginia Tech, discusses intercellular communication in the heart and how viral infection of cardiac cells can result in heart disease. 

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Jamie is Associate Professor at Virginia Tech, where his research aims to define cardiomyopathy at a subcellular level, searching for potential targets for therapeutic interventions to help restore normal cardiac function to diseased hearts.

Transcript of Episode 47: Jamie Smyth

Coming soon…

The Bioinformatics CRO Podcast

Episode 46 with Amar Gajjar

Amar Gajjar, world-renowned neuro-oncologist and chair of the Department of Pediatric Medicine at St. Jude Children’s Research Hospital, discusses emerging treatments for pediatric brain tumors such as medulloblastoma.

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Amar is Chair of Department of Pediatric Medicine at St. Jude Children’s Research Hospital. He is principle investigator on several clinical trials aiming to treat brain cancers like medulloblastoma and improve the quality of life for patients in recovery. 

Transcript of Episode 46: Amar Gajjar

Coming soon…

The Bioinformatics CRO Podcast

Episode 45 with Jill Reckless & Jon Heal

Jill Reckless, CEO and co-founder of RxCelerate, and Jon Heal, Head of In Silico Designs at RxCelerate, discuss the virtual biotech industry, outsourcing drug development, and using bioinformatics for target optimization.

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Jill is co-founder and CEO of RxCelerate, an outsourced drug development platform. Having completed her PhD at the National Heart and Lung Institute in London, Jill was an academic at the University of Cambridge until December 2011 before founding RxCelerate.

Jon is Head of In Silico design at RxCelerate. After completing a PhD at Imperial College London, he founded a computational biology-based drug development company Prosarix, which was acquired by RxCelerate in 2019. 

Transcript of Episode 45: Jill Reckless & Jon Heal

Coming soon…

The Bioinformatics CRO Podcast

Episode 44 with Adam Siepel

Adam Siepel, Professor of quantitative biology at Cold Spring Harbor Laboratory, discusses the applications of evolutionary genomics in anthropology, infectious disease, and cancer.

On The Bioinformatics CRO Podcast, we sit down with scientists to discuss interesting topics across biomedical research and to explore what made them who they are today.

You can listen on Spotify, Apple Podcasts, Google Podcasts, Amazon, and Pandora.

Adam is Professor of quantitative biology at Cold Spring Harbor Laboratory, where his lab studies molecular evolution and transcriptional regulation in cancer, infectious disease, anthropology, and agriculture.

Transcript of Episode 44: Adam Siepel

Grace: [00:00:00] Welcome to The Bioinformatics CRO Podcast. My name is Grace Ratley. I’ll be your host for today’s show, and today I’m joined by Adam Siepel. Adam is chair of the Simon Center for Quantitative Biology and Professor of Biology at Cold Spring Harbor Laboratory. Welcome, Adam.

Adam: [00:00:15] Thank you. It’s great to be here.

Grace: [00:00:17] Yeah, it’s great to have you. So can you tell us a little bit about the research that you’re doing?

Adam: [00:00:21] Yeah. So we do work in a variety of areas in genomics. My laboratory is completely a dry lab. We do only computational work, but we collaborate closely with a number of experimentalists and really try to stay as close as possible to data generation and biological questions. But we do have strong backgrounds in probabilistic modeling, algorithms, machine learning and related areas, and we try to bring those skills to bear on our work. We are interested broadly in questions involving evolutionary genomics, in particular evolution of gene expression. We are interested in demographic reconstruction of human populations involving both humans themselves and human interactions with archaic hominins such as Neanderthals and Denisovans.

[00:01:11] And we’re also interested in natural selection making inferences about the strength of natural selection, which parts of the genome are affected by natural selection, and natural selection on different time scales: ancient natural selection affecting primates and more recent natural selection affecting human populations, for example. And then, I should say also, we’re interested in applications not only in human population genetics, but also in cancer and agriculture and other areas. We’ve also done some recent work on COVID modeling, for example.

Grace: [00:01:43] Yeah. I saw a really interesting paper from your lab on COVID, looking at the influence of daylight savings time on immune response.

Adam: [00:01:53] That’s right. Yeah, it sounds like sort of a crazy connection. But one of my colleagues here, a senior scientist, Rob Martienssen, HHMI investigator, had a hypothesis that there could be an interaction between daylight savings time and seasonal patterns of COVID infection relating to the fact that at certain times of the year, people are most likely to be interacting with other people during their daily commute, exactly at the time of a nadir in immunity, which occurs around sunrise. And he had the observation that daylight savings time changing the clocks prolongs the period of time when the daily commute coincides with sunrise. And so we did some modeling to show that indeed, there seems to be a signal in the public data indicating that there is an effect and that infections could be reduced by eliminating daylight savings time. So that’s one of many reasons to eliminate daylight savings time.

Grace: [00:02:53] Yeah. I’m all for that research as long as I don’t have to wake up an hour earlier than normal.

Adam: [00:02:58] That’s right.

Grace: [00:02:59] Yeah. So can you tell me a little bit about how you get those data sets to predict how human genomes have evolved over time?

Adam: [00:03:09] Well, all of our research is based on data that has been collected by other groups. Much of it collected by Svante Pääbo group at the Max Planck Institute in Leipzig, Germany. And they have over many years now develop techniques for extracting DNA from fossilized bones. The techniques are quite sophisticated because if you’re not careful, it’s very easy to contaminate the ancient hominin DNA with modern human DNA. And so they’ve developed clean rooms and special DNA extraction techniques and special purification techniques. And then post-processing bioinformatics techniques to ensure that the DNA they’re sequencing really represents the ancient remains and not the modern humans who are handling the fossils. But we can’t claim responsibility or credit for any of those works. We’re consumers of the data that they have produced and made publicly available.

Grace: [00:04:10] Right. Of course, I didn’t know if it was some sort of reverse modeling like taking current human DNA to predict what the DNA looks like previously.

Adam: [00:04:19] Well, there’s some of that because what you get is kind of a noisy readout of the the DNA for the ancient remains. And then we have higher quality representations of modern human DNA. And then we try to model the processes that could have given rise to both of those samples. And that does involve sophisticated statistical methods for reconstructing ancestral DNA, as well as explaining the observed samples.

Grace: [00:04:48] Yeah. And so can you tell us a little bit about what you found in that research?

Adam: [00:04:53] Sure. It’s now fairly well known and well accepted based on findings that were developed over the last decade that there has been a genetic interaction between modern humans and Neanderthals. In particular human populations outside of Africa, including Europeans and East Asians show a signal of Neanderthal DNA at something like three percent of their genomes that traces back to Neanderthals. And the best explanation we have for that signal is that there was some sort of interbreeding between modern humans and Neanderthals, probably outside of Africa, after humans had migrated from Africa, something like seventy or eighty thousand years ago. And that signal persisted as these out-of-Africa populations spread across the globe.

[00:05:55] We came in already knowing these findings and already familiar with these findings. And we tried to develop a model that would jointly explain a number of ancient samples and a number of modern samples from around the world. And our goal was to see whether we could both explain this known pattern of Neanderthal-human interaction, but also possibly detect other signals of interest. And what we found interestingly, and this was published in 2016 in a paper in Nature that I jointly co-led with Sergi Castellano, who was then at the Max Planck Institute in Leipzig and has now moved to London. What we found was that there was a signal, surprisingly in the opposite direction of modern human DNA in Neanderthals. And this was something that hadn’t been reported previously. And the signal was quite subtle. And it was quite difficult to convince the community that it even existed.

[00:06:49] But we were able to convince reviewers of our paper, and it has since been supported by a variety of other analyses. And interestingly, this signal is not specific to out-of-Africa populations. It’s shared by Africans as well, and it appears to be much older. We’ve since in more recent work, dated it to somewhere around two hundred fifty thousand years ago. And so that suggests that there was an earlier integration event that left a signature in the opposite direction from modern humans to Neanderthals, and it affected all human populations. So it probably occurred in the ancestor to all modern humans. Furthermore, that’s interesting because it must have predated the migration of humans out of Africa. So it seems like there was a group of early modern humans that migrated out of Africa interacted with Neanderthals, leaving this signature in Neanderthal DNA that we’ve detected.

[00:07:52] And then that group of modern humans either just went extinct or ended up being absorbed back into human populations in Africa before a second migration out of Africa seventy or eighty thousand years ago. So anyway, it suggests not only another interaction between modern humans and Neanderthals, but one that’s much earlier, and it paints a picture of multiple migrations of modern humans out of Africa. And only the more recent cases led to the current populations that we know of today outside of Africa.

Grace: [00:08:28] Yeah, that’s so fascinating. I feel like bioinformatics is already such an interdisciplinary subject. I mean, taking together biology and computer science, and then you take it a whole step further and add anthropology in there. Do you work closely with people in anthropology or studying human history?

Adam: [00:08:48] Yeah, I have not worked directly with anthropologists myself. Although we did have an anthropologist collaborator on the paper in Nature, although we worked more closely with Sergi than with me. But there is a lot of interest across the field at this intersection between genetics and anthropology and Svante Pääbo and David Reich and others have been quite proactive about interacting across fields. I attended a meeting here at Cold Spring Harbor a few years ago that was organized by co-organized by David Reich. That was a group of geneticists and a group of anthropologists together discussing these issues, and it was fascinating. But it’s not something that’s really in the center of my own research.

Grace: [00:09:29] You seems to have a very broad reach with your research. It’s great. Yeah. So you started out doing computational modeling and phylogenetic modeling in HIV. Can you tell us a little bit about that work?

Adam: [00:09:41] Yeah. So this was my first job actually straight out of college. I hadn’t gone to graduate school yet. Through a friend of mine who had been an undergraduate with at Cornell, found out about this opportunity to work at Los Alamos, doing HIV sequence analysis. And it was interesting to me for a number of reasons. I had done an undergraduate degree in agricultural and biological engineering. And so I had been interested for a long time in sort of the intersection between mathematical modeling and questions in biology. But I had never dealt with DNA sequence data before, and I had never dealt with phylogenetics or evolutionary reconstruction. And when I was exposed to those fields, I just found them fascinating. I mean, they resonated with me in a whole variety of different ways. I’ve always been interested in reconstructing the past.

[00:10:28] I’m interested in random processes. I’m interested in computer algorithms. I’m interested in evolution. And so all of these interests sort of came together in this fascinating area of using phylogenetics. And then that work also had an epidemiological component. We were building phylogenetic trees to describe HIV sequences, but then we were making use of them to understand the spread of HIV across the globe because we were seeing different strains emerge in different regions of the world. And then we began to see interactions between these strains and the production of recombinant strains of HIV. So my first scientific paper was actually on an algorithm that I developed, a very simple algorithm, to detect recombinant strains of HIV, which at the time was a kind of a new idea and something that was of great interest in the field.

Adam: [00:11:23] My first experience was very exciting publishing a scientific paper. I think I was 23 years old and was able to publish a paper that established researchers in the HIV field were excited about, and I got to present at meetings and so on. And after that I was hooked. I was hooked on science, I was hooked on computational biology and I was hooked on evolutionary genomics.

Grace: [00:11:45] Do you still keep up with emerging HIV research and things like that?

Adam: [00:11:49] I haven’t participated in HIV research since that time. I moved on to other questions. Although I have to say I got interested again recently in this question of recombination and viruses with the emergence of COVID-19. And reread some of those old papers and including my own old work on detecting recombination in viruses because there was some discussion about the role that recombination might have played in the emergence of SARS‑CoV‑2 in human populations. But that’s my only experience in that area in the last 25 years or so.

Grace: [00:12:30] Yeah. So I guess speaking of the pandemic, from an evolutionary standpoint, do you think we need to worry about new variants and things like that?

Adam: [00:12:41] The basic evolutionary fact is the probability of emergence of a new variant should be proportional to the number of viral replication events, which is going to be proportional to the overall number of cases. And so we need to get the number of cases down. And the best way to do that is through vaccinations. It’s been extremely discouraging to me to have these effective vaccines, more effective than anyone could have hoped, and see people reluctant to use them. So I think we just have to keep hammering on the vaccination efforts. They need to be available across the entire world, not just in rich, first-world countries. We need to push really hard on getting access to them, convincing and incentivizing people to use them.

[00:13:36] Ultimately, I think new variants will emerge. We will develop over time an increasing sort of baseline resistance for most people in the world who will eventually be exposed. And I think the pandemic will eventually reduce itself to a baseline level. But I think the virus will be endemic and we’ll have to adjust to it being part of life. I’m optimistic that with increased baseline resistance, increased vaccination, increased ability to provide new vaccines quickly and efficiently, that we won’t be brought back to our knees by emerging variants. But it’s difficult to say for sure what could happen as new variants emerge.

Grace: [00:14:29] I always like to hear the different perspectives of people in different fields of science on the pandemic. I think the evolutionary take on it is very interesting. You also did a study in bats, on the evolution of bat immunity and things like that.

Adam: [00:14:46] That’s right. Yeah, we’ve gotten very interested in comparative genomics of bats, in part because of their connection with SARS-CoV-2. But for other reasons as well. In fact, our initial work on bats has been funded by our cancer center here at Cold Spring Harbor. Because bats are remarkably resistant to cancer and we’ve been trying through DNA sequencing and comparative analysis to shed light on the genetic underpinnings of both bat specific immune responses, bat specific cancer resistance and longevity of bats. Bats are extremely long-lived mammals for their body size. If you plot body size versus lifespan in mammals, you see a general proportionality. But bats are an outlier. They live much longer than other mammals of similar size, such as mice. Bats can live 35-40 years or more.

[00:15:42] We’ve been doing DNA sequencing and analysis. We have an initial preprint out on our findings. We’ve found some interesting things in both immunity and cancer, in particular a massive contraction of the IFN1 locus. And a strong enrichment among apparently positively selected genes for tumor suppressors and DNA repair genes. And we’re in the process of working closely with experimentalists to begin to test the actual molecular basis of some of these differences between bats and other mammals. And we’re also in the process of applying for grants from the NIH on this topic.

Grace: [00:16:22] That’s so cool. Because I guess I wanted to ask a little bit about the importance of evolutionary biology in the study of cancer, because I wasn’t necessarily sure how those two topics connected. So that’s a really interesting take on comparative genomics and looking at how that immune system has influenced their susceptibility to different cancers. And bats have a reduced immune response. They don’t have a very active immune system, is that correct?

Adam: [00:16:51] Yeah. They seem to be able to tolerate viral infections without having a very powerful immune response. And it’s interesting because when you look at what makes humans sick when they become infected by SARS-CoV-2 or other viruses, it’s often an overly powerful immune response that makes them very sick. And so in some cases, it seems that viruses are killing us, not because our immune response is inadequate, but because it’s too powerful. And one hope is that we can learn something from bats in the way that they’re able to keep from getting sick from these viruses and yet not have an overly powerful immune response that ends up harming them more than the virus itself. Yeah. So that’s one of our interests in this area.

[00:17:44] Of course, it’s also interesting to just understand the dynamics of zoonotic transmission and the way in which bats are harboring viruses and then transmitting them to people. The fact that bats seem to be able to tolerate such high viral loads does seem to be essential to their role as a reservoir for viruses that get transmitted to humans. And so understanding their viral tolerance is also important and interesting in that regard.

Grace: [00:18:16] Yeah. So I mean, evolutionary biology is kind of a pretty popular science topic. So what do you think are some misconceptions that people in the general public have about evolutionary biology?

Adam: [00:18:30] One misconception is understanding the diversity of selective forces that have influenced humans. People tend to think in conventional terms about the strongest humans being the ones that propagate, you know, the ones that are least likely to be killed by predators and that sort of thing. And undoubtedly avoiding predators was a source of selective pressure on humans. But there are many others that I think tend to be underappreciated. One of them is infectious disease. I mean, humans have been enormously shaped by infectious disease. And one of the strongest selection pressures on us is the resistance to infectious disease. The pandemic is helping make this issue more clear. But I think in general, we tend to have forgotten a lot about infectious diseases because they play much less of a role overall in modern life than they have in the past.

[00:19:28] Another really important selective pressure is sexual selection. The choices people make about who they mate with for various reasons. And then there are very strong selective pressures that influence reproduction in a way that humans have no choice over. So, for example, sperm competition individual sperm cells competing with one another to fertilize an egg. So there are many, many levels at which selection acts. And I think when people just think about a caveman dodging a mountain lion or a bear, they’re only getting at a very small sliver of the diversity of selective forces that have influenced human evolution.

Grace: [00:20:14] Yeah, that’s true. There are some really interesting selection events. So after you worked in Los Alamos, where did you head after that?

Adam: [00:20:22] Well, I was working in Los Alamos in the mid 90s. And I had an engineering background and I had a lot of interest in developing computer software. And at the time, I felt that my interests lay more in the software development area than in the scientific research area. And it coincided with a time where there was a lot of opportunity for software development in bioinformatics. A lot of companies were creating bioinformatics groups. A lot of people were developing and either selling or making publicly available bioinformatics software. And so I took a job at a group in Santa Fe, New Mexico, called NCGR, National Center for Genome Resources that was doing a lot of software development. I went there and I worked for about 5 years as a software developer and learned a lot about software development and then kind of came to the conclusion that I wanted to get closer to the science.

[00:21:20] And after many years of putting off going back to graduate school, I decided I really needed to bite the bullet and get my PhD. I was sort of a reluctant academic, I have to say. At the time, I was of the mindset that I could teach myself anything I needed to know. But I finally decided that there was value in getting my PhD and diving back into scientific research. So I left software development, became a full time PhD student and went to Santa Cruz, California, to join David Haussler Laboratory. And from that point on, I have plunged myself into the world of comparative genomics, population genetics, evolutionary modeling and so on.

Grace: [00:22:01] Yeah. And despite that reluctance to going into academic science, you stuck with it after your PhD because you went and became a professor at Cornell and now Cold Spring Harbor. Can you talk a little bit about that decision? How did you change your mind?

Adam: [00:22:16] I actually had not planned to go into academia. I wasn’t sure what I was going to do. But it was an exciting time, the early 2000s for academic computational biology. There were a lot of opportunities emerging, a lot of new departments, new research centers. And in my third year as a PhD student, I had been working with Rasmus Nielsen, who’s now at UC Berkeley, on a book chapter project. He was editing a book and I was writing a chapter with my advisor. And he sent me a job ad at Cornell and I read this job ad and it just sounded like it was written for me. I mean, they were looking for someone who had exactly the sort of expertise I had. And, you know, I had been an undergraduate at Cornell, so I had a lot of affection for the place.

[00:23:07] Coincidentally, I also was considering moving closer to family. My family’s from upstate New York and Cornell is in upstate New York, and I had two small children and we were getting tired of putting them on planes every time we wanted to see family. So I said, what the heck? I’ll apply to this job. I applied and I got the job. So I said, well, you know, I never really planned to be an academic, but this sounds like fun. It sounds like a great opportunity. I love what I’m doing. This is an opportunity to keep doing what I’m doing. And I took the job and never looked back. I’ve really enjoyed academic work since then and have been able to make it work, been able to keep the lab funded and keep publishing papers and keep recruiting students.

[00:23:49] And I think I’ll just keep doing that as long as I can. But it was a different time. I mean, I mentor a lot of my own graduate students and postdocs in their job searches. And I think the job market is much more competitive now than it was then. There was a lot of opportunity in computational biology in the early 2000s, and I benefited from being in the right place at the right time. Sometimes I see the job searches we carry out now, and I wonder if I would have even gotten an interview for some of these jobs.

Grace: [00:24:18] Yeah. Academic science is a very competitive space these days. But there is such a strong need for bioinformaticians and computational biologists. So I mean, there’s a lot of job security in that, but maybe academia is a lot harder.

Adam: [00:24:34] Yeah. I think there are more industry opportunities now than there were at that time. And, you know, the combination of the competitive academic job market and the opportunities in industry means that a lot of young trainees are going into industry, which I think is great. I have a number of recent postdocs from my lab who’ve taken industry jobs and are very happy in them. But, you know, the pendulum tends to swing from one side to another on these things. And I wouldn’t be surprised if in a few years the supply and demand dynamics have changed and things open up in academia again.

Grace: [00:25:08] Certainly. And how have you seen bioinformatics and computational biology as a field evolve in the last few years?

Adam: [00:25:18] One change is just, as I mentioned, a swing toward more activity in research and industry. Another change that I’ve seen in my time in computational biology is just a general shift toward embracing the biology side of the field. They need to ask good biological questions, they need to engage with the data and people not being satisfied with just taking whatever the latest algorithmic or machine learning advances and applying it to a biological data set. I think when I started in the field in the early 2000s, there was a lot of that. There were a lot of people doing computational biology who weren’t that interested in biology and didn’t know that much biology. They were just taking off-the-shelf computational methods and applying them to biological questions in a not very imaginative way.

[00:26:10] And I think over time, people have really realized that in order to do computational biology well, you have to engage with the biology. It’s not enough to just have a computational hammer and look for nails. You have to really think imaginatively about biological questions and how computational methods can be used to address them. And about the interaction between computational methods and experimental methods. About how experimental methods can lead to hypotheses that can be tested computationally and vice versa. Computational methods can generate hypotheses that can be tested experimentally. That feedback between computation and experiment, I think is extremely important and has become more pervasive in the field.

[00:26:54] I think the field is also just bigger and more competitive. Early on, there were really just a handful of people who had this joint background in computer science and biology. And if you were one of those people, then you could sort of write your own job description. It was relatively easy to find a job in the field. Now there are many, many people who have those backgrounds. There are people emerging from PhD programs in bioinformatics and computational biology. There’s a lot more awareness of these questions in biostatistics departments or biophysics departments. It’s just a much more established and competitive academic field.

Grace: [00:27:39] Do you think you would have chosen the same path if you had graduated in bioinformatics today?

Adam: [00:27:46] I really don’t know. I mean, I think I was attracted to the field being so new. And maybe I would feel today that it was too established and I would look for something newer and more niche. But it’s hard for me to say. I also think it’s possible that if I were finishing my PhD now that I would end up in an industry job rather than in an academic job just because of the dynamics of the field at the moment. But it’s always hard to ask these counterfactual questions.

Grace: [00:28:19] True, true. So given the hyper competitive job market for positions and bioinformatics, can you maybe give advice to people who want to enter that field? Like what sorts of skills are most important today?

Adam: [00:28:37] Yeah. I guess, I think it’s true that a graduate from a bioinformatics program who’s interested in this field needs to be fluent in data science and machine learning, basic statistics. But I think that those things are necessary but not sufficient for success in the field. And I think what really will push a person over the edge is also really thinking like a scientist, not just like an engineer. So developing a good taste in problems, developing a nose for questions that can be effectively addressed using computational methods, developing a fluency in the biological technologies and biological questions of interest, the ability to interact closely with experimentalists. I think these are the things that push a person over the edge from being just a data scientist to being a computational biologist who can lead the way in the scientific side of the field.

Grace: [00:29:41] It’s very good advice. So tell me a little bit about you. Like, who is Adam the non-scientists? What do you do outside of research?

Adam: [00:29:49] Well, I have two kids. My daughter just started at the University of Rochester. So I’m adjusting to going from two kids at home to one kid. I live on Long Island in Huntington, New York and live in an old Victorian house and spend a lot of my time fixing that up. And I like to do a lot of cycling and some hiking and spend as much time as I can outdoors. That’s probably a pretty good summary.

Grace: [00:30:16] Yeah. I actually saw in your Twitter that you were planning on heading to Iceland. Did you make it out there?

Adam: Yeah, we did.

Grace: Nice. Yeah, I was just there a couple of weeks ago.

Adam: [00:30:27] Ok. Yeah, we really enjoyed it. We had a fantastic trip. It’s a beautiful place and it felt like the right sort of first trip out of the country after COVID. Relatively safe and controlled.

Grace: [00:30:39] Yeah, that’s excellent. Yeah. Actually, Iceland is a really, probably a very interesting area to study because it’s so isolated and they have a huge dataset. Haven’t they sequenced everybody in Iceland?

Adam: [00:30:52] Yeah. The studies by deCODE have been extremely influential in a variety of different ways, both for association studies and also for studies of rates and patterns of human mutation, which they’re able to trace in great detail, taking advantage of their genealogical databases and pedigrees. So, yeah, it’s been very important in human population genetics. It’s also interesting to look at Iceland from an ecological perspective. I think the largest land mammal was the Arctic Fox in Iceland when Scandinavians arrived and began bringing agricultural animals. So there’s a very short history of large land mammals there. And then there have been interesting events like the introduction of the Icelandic horses and then subsequent genetic isolation, of those horses. And it’s interesting to see the way they have been shaped by the Icelandic landscape and climate, as well as by human selection. But yeah, it’s a fascinating place for questions in evolutionary biology. Certainly.

Grace: [00:31:59] Certainly. And yeah, Iceland horses are really interesting. They had such strict laws that if an Icelandic horse was taken out of Iceland that it couldn’t be brought back into the country. It was just really interesting. And with humans, they have an app. It’s like a dating app where you can check and see if it’s okay to date somebody based on your familial relationship to them.

Adam: [00:32:24] Ah, to see whether you might be related, yeah.

Grace: [00:32:26] Yeah. You put their name in and I think the generally accepted okay line is like fourth cousin or something like that.

Adam: [00:32:33] I see. Well, amazing.

Grace: [00:32:35] Yeah, it’s a really interesting country. Yeah. So as we wrap up the episode, do you have any other any final thoughts on the future of bioinformatics?

Adam: [00:32:46] Well, I guess the future of bioinformatics, I think it’s an open question whether bioinformatics will remain a distinct field. I think that to some degree, the tools of bioinformatics are being absorbed by broader biological sciences. They’re just becoming part of the toolkit of doing biology. And I think in the future, biologists will need to be much more fluent in computational methods and the use of machine learning and the use of powerful computers. And we may not think of it as a distinct field. It may just become part of being trained to do biology. And I think that’s okay. I think often new fields emerge at the interfaces of other fields, and they may or may not remain distinct. They may be absorbed over time, and I think that’s okay. I’m personally very excited to see quantitative methods and computational methods become so central in biology.

[00:33:50] You know, our center at Cold Spring Harbor, it’s called the Science Center for Quantitative Biology. It has begun as kind of a distinct group of investigators doing developing quantitative methods. But increasingly we’re being absorbed by the broader scientific community at Cold Spring Harbor. And the talks when we gather at our annual symposium or some other event to talk about our research. The talks from the quantitative biologists are beginning to involve more experimental biology and more collaboration with experimentalists. And then conversely, the talks by the experimentalists are beginning to incorporate more data analysis and quantitative methods. And I think the logical conclusion of this process is that we probably won’t be a distinct group anymore. We’ll all just be biologists using whatever tools and techniques are available, a combination of experimental and computational tools and techniques. So I guess that’s what I think about the future of the field. It’s dying, and that’s okay.

Grace: [00:34:56] It’s dying, and that’s a good thing. Fantastic. Well, thank you so much for joining me today, Adam. I had a great time listening to your thoughts on evolutionary genetics.

Adam: Yeah, thanks, Grace.