Showing posts with label COVID-19. Show all posts
Showing posts with label COVID-19. Show all posts

Wednesday, May 26, 2021

Origins of SARS-CoV-2

Here are some links to helpful articles about the "lab-leak" hypothesis:

Nicholas Wade (link), Yuri Deigin (link), Rossana Segreto and Yuri Deigin (link)

Andersen et al. (link), Yangyang Cheng (link), Glenn Kessler (link), Matt Yglesias (link)

Daniel Engber (link)

My favored hypothesis has been that the Covid-19 pandemic began in the same way the original SARS pandemic did, in a "spillover" of SARS-CoV-2 from a natural reservoir of the virus in bats, by way of an intermediate host.

Along with much of the scientific and political world, this week I've been reading, and updating my likelihood of the alternative hypothesis, that the pandemic began when SARS-CoV-2 escaped in an accident, from a lab at the Wuhan Institute of Virology (WIV).

The problem with the spillover hypothesis is that, unlike with the original SARS, there is very little supporting evidence.  Rather, what we know about SARS-CoV-2 indicates that the virus appeared in humans in a single event. There is very little diversity in sequences of early virus samples.  It was not strongly associated with a "wet market" and there is no known intermediate host.

When the virus appeared in humans, it was already well-adapted to growth in them, with an S protein that fits its human receptor (the ACE2 receptor protein) very well, as well as a furin-cleavage site that is known to facilitate viral entry.  The furin site is essential for infectivity in humans.

With respect to the escape hypothesis, it did not help that Donald Trump supported this idea, suggesting the sinister possibility that release not only explains the pandemic, but that it was deliberate.  Since Trump has lied repeatedly about nearly everything, his statements actually decreased (at least for me) the likelihood that this hypothesis is correct.  Add to that the fact that so many right-wing politicians are looking to pick a fight with China.

I do not put much weight on vague reports that WIV staff were hospitalized with flu-like illness in the fall of 2019, including other reports that samples from them were tested for SARS-CoV-2 and came back negative.  The former reports come from "intelligence sources" who were motivated to support Trump's contention, and the latter from official Chinese government sources equally motivated to keep things under wraps.  They are possibly true, but it is impossible to evaluate their accuracy now.  

I also do not think much of the idea that, as David Baltimore reportedly said, the furin site is a "smoking gun".  Acquisition of such a sequence could easily have happened naturally, by recombination.  The site's codon usage does not provide convincing evidence that it was engineered.  Indeed, similar sequences that might have been recombined to give the furin site would perhaps also have been out-of-frame originally, and recent acquisition hasn't given enough time for codon optimization to the human host.  Also, if you were engineering a site, you would not likely have chosen this sub-optimal sequence over known examples that work better.

Modern methods do not leave signs in the sequence (such as convenient restriction sites for assembling a whole genome from DNA clones of parts of it) that would be unmistakable "signs of engineering" for older techniques.  So the frequent claim that the sequence doesn't show such signs isn't dispositive either.

What I find most troubling is the reporting that WIV (as well as labs elsewhere) were doing manipulations with parts of the S gene (the RBD).  They built hybrid viruses with S genes whose product was known to bind well to human cells.  These were done not just with pseudo-typed lentiviruses (which would be safe), but with bat coronavirus backbones (which is definitely unsafe).  Incredibly, some labs even added furin cleavage sites.  Multiple labs did such experiments, in China, in the US and Japan, and in other places.

Adaptation experiments were also carried out, by serial culture in human cells and in "humanized" mice bearing Hu ACE2 receptor.  Richard Ebright is correct to insist that all these experiments were exceedingly irresponsible.

A second problematic aspect is lab safety.  Shi Zheng-Li, the PI on these projects at WIV, stated that their work was carried out at BSL-2 and BSL-3, and Wade seizes on this to claim repeatedly that dangerous experiments were carried out at BSL-2.  If this is true, then that would also be very very bad, and even in the absence of any other confirming evidence, would weight the scale heavily in favor of the escape hypothesis.

A third problematic feature is the publication of a viral sequence closely related to SARS-CoV-2 (the one most closely related) just after the beginning of the pandemic, by Shi and collaborators.  This sample was clearly from around 2013. It is reported that the sequence file naming scheme indicates it was sequenced in 2017 and 2018.  Their attempt to hide this and pretend that it was only sequenced recently, is also troubling, if this is true.

Even if the relationship between the US and China were less antagonistic, I do not imagine that the Chinese government would ever allow investigators to access the original notebooks and sequence databases of the WIV investigators.  So at the end of the day, we shall probably never know the truth.  All we are left with is likelihoods that are heavily weighted by our "priors" about the two hypotheses.  

The real bottom line here is that even if the pandemic started by spillover, it certainly could have started by lab escape.  We should act on that, to ensure it doesn't happen again.

Update:  

Just to be clear though, the post above is my analysis of the likelihoods for two different hypotheses of what might have happened.  

Florian Kramer advises patience.  It took more than one year to find the relevant evidence for SARS.  For some other viruses, we have still not found the source of the spillover.  I think his priors for lab escape are too low, but still, this thing is unresolved.  It might change in the future if evidence supporting spillover becomes available.

However, if the leak theory is what actually happened, the evidence is going to be buried very deep.  Plus, the Chinese will never allow such an investigation.  All that pressing for a scientific investigation will do is lead to China-bashing, which is not helpful.  And the idea that the American intelligence community will sort this out is laughable.

What matters most is for us to accept that it could have been a leak, and change our attitude about whether research of that type should continue.

Wednesday, June 24, 2020

COVID-19 data analysis

I've been working on a project to analyze the data on COVID-19 cases and deaths (GitHub repo). The data is aggregated from public health departments by some folks at Johns Hopkins.  Their dashboard is here, and the data are in this directory on Github.

Here is the help page for the covid project (it is the same for most of the scripts):
> python3 one_state.py --help
flags
-h  --help    help
-n   <int>    display the last n values, default: 7
-N   <int>    display N rows of data: default: 50
-c  --delta   change or delta, display day over day rise
-d  --deaths  display deaths rather than cases (default)
-r  --rate    compute statistics
-s  --sort    (only if stats are asked for)
to do:
-u   <int>    data slice ends this many days before yesterday 
-p  --pop     normalize to population
example:
python one_state.py [state] -n 10 -sr
And here is the output (today) for that example:
> python3 one_state.py SC -rs
               06/17 06/18 06/19 06/20 06/21 06/22 06/23  stats
Charleston      1264  1403  1554  1728  1836  2044  2251  0.094
Oconee            95   100   105   110   136   154   142  0.083
Pickens          348   367   429   464   499   529   570  0.083
Calhoun           47    48    58    62    69    73    74  0.082
...
total          20556 21533 22608 23756 24661 25666 26572  0.043
The statistic is a linear regression on cases, normalized to the mean of the values, and then the counties in my state (SC) are sorted according to the result.  Charleston is my county, and unfortunately, it is the county with the highest rate of growth of cases in the state.  Currently in the US, the top 20 counties are:

> python3 us_by_counties.py -rs -n 4 -N 20
                   06/20   06/21   06/22   06/23  stats
Thomas, KS             0       0      10      12  0.836
Hot Spring, AR        53      53     138     226  0.514
Holmes, FL            47      47      58     121  0.341
Jim Wells, TX         22      27      34      46  0.245
Brewster, TX          24      24      39      45  0.236
Erath, TX             44      44      44      85  0.227
McDonald, MO         170     366     371     403  0.215
Sharkey, MS            9       9      13      16  0.213
Blanco, TX            14      14      22      24  0.205
Newton, TX             6       8      11      11  0.2
Aroostook, ME         11      17      19      21  0.188
Tehama, CA            34      34      53      54  0.181
Sioux, ND             12      12      19      19  0.181
Okfuskee, OK           7       7      11      11  0.178
Bourbon, KS            9       9      14      14  0.174
Lawrence, MO          11      13      13      19  0.171
Harvey, KS            13      13      20      20  0.17
Pontotoc, MS          93      93     128     146  0.169
Letcher, KY            8       8       8      13  0.162
Live Oak, TX          10      10      15      15  0.16

I chose n = 4 so the output would be formatted correctly for the blog post.

As with any large dataset, there are some problems to work through, which are not solved perfectly yet.  Also, I've focused more on the U.S. lately, so scripts for world data haven't been updated yet either.

What I got interested in and want to show is the generation of maps of the US by states or counties, or one or a few states by counties, where the fill color is based on, for example, the growth rate of cases.  Here is the US by states.



I haven't generated the color bar horizontally yet, I just cut it out and rotated it, so the writing is rotated as well.

This type of map is called a choropleth map.  I stumbled across a python tool for generating maps.  It's part of the plotly library.  It is as simple as 

fig = px.choropleth(
    df,
    locations=abbrev,
    locationmode='USA-states',
    color=st,
    color_continuous_scale='Plasma',
    scope="usa",
    labels={'color':'growth'})
fig.show()

The details are slightly complicated, but not bad. 

df is a pandas data frame that maps states by two-letter abbreviation to the corresponding statistic.
df = pd.DataFrame(data={'state':abbrev, 'value':st})
You need GeoJSON data for a county map (the states are already known to plotly.express).  That data file is available from them.

The colors are mapped to the statistic st as read from the data frame.  The last line of the call to px.choropleth assigns the title to the color bar.

The script for the US states is here and for the counties it is here.

This is the state of South Carolina today.  These colors are an attempt to make the positives pop out more.



There's much more to discuss.  I have always wanted to make a map of the US with my road trips plotted on it, something like this.  For that we need to talk about GeoJSON data and how to obtain it, as well as the Albers projection that is used in making maps.  It turns out that the standard methods from plotly have a significant limitation and I had a really weird bug in my code that I eventually figured out.

Finally, we'll need to find how to generate the data for each individual trip to overlay on the map.  That's all for later.



Saturday, May 9, 2020

Reconstruction of SARS-CoV-2

A few days ago a paper was published in Nature after rapid review:

Rapid reconstruction of SARS-CoV-2 using a synthetic genomics platform.

It's quite amazing.  For starters, the work was completed in less than 6 weeks, since the first genomic sequence of the SARS-CoV-2 virus was released on Jan 10 and the paper was submitted on Feb 22.  Most of that time was preparation of the DNA fragments.

9 DNA fragments

The first step was to PCR amplify DNA segments spanning the genome.  This was done by RT-PCR.  Usually one verifies the sequence has not been mutated in the PCR reaction (I don't see this in the paper).  

Alternatively one can just order synthetic DNA of < 8 kb these days (with a 10 day turnaround)!  link NPR story

Cloning

Some DNA fragments are unstable when cloned into E. coli.  So they used YACs (yeast artificial chromosomes).  An additional advantage is that all 8 fragments can be assembled into the final product in one step!  (TAR-cloning)


In vitro transcription

The clones are made with a promoter for phage T7 RNA polymerase upstream and poly A tail downstream followed by a restriction site for Pac I.  YAC DNA was prepared, cut with Pac I and transcribed with T7 polymerase.

Transfection

The transcript was then transfected into mammalian cells, together with mRNA for one viral protein.  The system is a bit more complicated than that:  there are two cell types.  But it seems easy enough, and the result is that the transfected cell system produces virus.



The virus can then be cloned (classic terminology, progeny from a single individual produced) and grown in culture.  Simply amazing.

Attenuating mutations from viruses for the live-attenuated vaccine for SARS-CoV can be assembled into live SARS-CoV-2 and tested for growth properties and attenuation, and then investigated as vaccines.




Friday, May 8, 2020

Remdesivir

You've probably heard about remdesivir, it's all over the news.  Tony Fauci was on TV being excited about a not very exciting study the other day that "showed promise."

I've been quite interested in this drug ever since I read in NEJM that the first patient to be diagnosed as SARS-CoV-2 positive in the US ("Snohomish"), was treated under an EUA with remdesivir when they thought he was going south, and the next day, he got better.  Much better.  Supplemental oxygen was discontinued that day.  There is plenty of anecdotal evidence on the web in addition to Snohomish.  I read a piece about an ER doc the other day but I can't find it now.  The patient said "that magic juice works, Doc."

Remdesivir has a weird name.  Many drugs do. In this case, it's a result of a naming convention that antivirals end in -vir, monclonal antibodies in -mab and so on.  It's a very useful convention because one can immediately tell the class of a drug from its name.  There's a whole list of them here.  I don't know that remdes- is significant itself.

Like many antiviral drugs, remdesivir is a nucleoside analog.  Here it is compared with ATP (remdesivir on the left, adenosine triphosphate on the right).



Remdesivir was initially developed for Ebola and Marburg viruses (filoviruses).  It has broad antiviral activity (in vitro) for other virus families including paramyxoviruses and coronaviruses such as the first SARS-CoV.  These viruses are all RNA viruses, and remdesivir inhibits the enzyme that replicates the genome, which is an RNA polymerase.

As part of its clinical testing for Ebola, the safety profile in humans is known to be good, although there are some things to watch for.

As with all antivirals, it is likely to be most effective if given early in the disease.  The problem with that is that there doesn't seem to be much of it.  The maker, Gilead Sciences, just contributed its entire stockpile of 1.6 M doses to the federal government, and they are busy sending most of it states that voted for Trump (I'm not kidding).

Remdesivir is apparently hard to make.  Here is an article about some chemists who were able, with a lot of effort, to synthesize one gram of it.  Although you can take that with a grain of salt (or a shot of tequila).  This story has twists and turns in it.

Gilead

So Gilead is a company started in the late 1980s, the time when many people including my former professors John Abelson and Mel Simon had a startup to find drugs active against reverse transcriptase.  It is named after the Balm of Gilead.

Gilead developed a couple antivirals in the 1990s, and then (small world department) acquired a company called NeXstar Pharmaceuticals in Boulder (this is Larry Gold's company, another colleague from the phage T4 world).  If I'm reading wikipedia right, NeXstar was their key acquisition because they obtained a sales and marketing ability for Europe and other markets.

In any event, Gilead is the fish that keeps eating other fish its size or even a little larger, exactly the right ones, and then gaining weight.  Big enough to give away a lot of money.  According to the wikipedia article, "Charitable donations to HIV/AIDS and liver disease organizations totaled over 440 million in 2015."

One of the companies they acquired had developed sofosbuvir, an antiviral for Hepatitis C virus.  They paid 11 B and yet Forbes calls it ""one of the best pharma acquisitions ever".

You may have heard about these treatments.  HCV (hepatitis C virus) is essentially a death sentence.  Gilead's drug will cure you, but it'll cost you your house.

the cat story

So finally, there's a weird tie-in with cats, which I read this morning in the Atlantic.  (In fact, I've been so impressed with them, especially recently, that I purchased my first magazine subscription since Mad magazine).

Apparently remdesivir has a cousin, a closely related drug called GS-441524, also developed by Gilead.  It was discovered by a research scientist working on the disease FIP, that GS-441524 cures cats of this fatal GI illness, which is caused by a coronavirus.  Apparently, in spite of the pleas of legions of cat-lovers, Gilead will not try to license GS-441524 for this purpose, apparently because
any adverse effects uncovered in cats might have to be reported and investigated to guarantee remdesivir’s safety in humans.
So there is a shady black market in China for GS-441524.  I kid you not.  Read the story, it's great stuff.

Summary

This is just my two cents, which means little.  I am highly skeptical that a vaccine can be successfully developed in 18 months, let alone 6.  As I said in another post today, the optimistic experts say "maybe we will get lucky."

But I am very optimistic about therapeutics and particularly remdesivir, given early.  The thing about COVID-19 is that the lung damage is so severe.  Even if you survive it takes a long time to heal.  And antivirals always work best given early.  See oseltamivir (another Gilead drug).

DNA and RNA vaccines for SARS-CoV-2

People ask about the prospects for a vaccine for SARS-CoV-2.

The NYT had a good article about it about a month ago.

The history of vaccine development is a series of long drawn-out and frequently failed attempts.  The idea that we throw everything at this problem and it's just going to work, is pie-in-the-sky.  The optimistic experts say "maybe we will get lucky."

The proposals that would move the fastest are for DNA and RNA vaccines.  These are vaccines that would contain (for example) the gene for the Spike surface protein of the virus, with accessory elements giving high expression.

They would either be mRNA (modified to be resistant to breakdown) or DNA.

But you should be aware that no DNA or RNA vaccine has been licensed for humans, period.  There is one for West Nile Virus in horses.  Many have been tried but they frequently don't elicit neutralizing antibodies.

One problem is efficiency.  The nucleic acid just doesn't get into cells very well.  People are trying gene "guns" to improve efficiency:  the nucleic acid is bound to small metal microbeads, and propelled into your arm with compressed helium.

Moderna is an RNA vaccine that is in ongoing Phase 1 clinical trials, and has been approved to start Phase 2 even though the results from Phase 1 aren't known yet.  Phase 3 is expected to begin in early summer.

Summary

Phase 1:  Small N.   Is the vaccine grossly toxic or relatively safe?
Phase 2:  Medium N.  Is the vaccine effective?
Phase 3:  Large N    Is the vaccine very safe and also effective?

Analysts are even wondering where they are going to get all the vials they will need to put the vaccine in, and the syringes and needles for injections.  It's a real problem.


Thursday, May 7, 2020

Links to recent posts inspired by the pandemic

Covid-19 pathogenesis and epidemiology:
link Unsupported claim of increased pathogenicity (2020-04-23)
link Covid-19 deaths by county (2020-04-29)
link South Carolina cases (2020-04-30)
link Covid-19 CFR (2020-05-01)
link SARS-CoV-2 transmission (2020-05-03)
link Covid:  good news, bad news (2020-05-05)

SARS-CoV-2 virology:
link Phylodynamics, NextStrain, and "Snohomish" (2020-04-30)
link Rapid reconstruction of SARS-CoV-2 using a synthetic genomics platform

General virology:
link Introduction to animal viruses
link Measles virus

Vaccines:
link Overview of vaccination (2020-05-07)
link DNA and RNA vaccines for SARS-CoV-2 (2020-05-08)
link Virus attenuation (2020-05-18)

Therapeutics:
link Remdesivir

Biochemistry: 
link Covid tests:  PCR (2020-05-06)
link Covid tests:  LAMP + CRISPR (2020-05-06)
link Sensitivity, specificity and Bayes (2020-05-06)
link DNA sequencing:  classical methods (2020-05-08)
link Chromatography (2020-05-10)
link Electrophoresis (2020-05-10)

Literature:
link Lewis Thomas on Streptococcal pneumonia (orig FB 2020-03-29)

Stupid stuff debunked here (links):
link Lyme disease is a U.S. government-developed bioweapon (false)
link Mouse retrovirus associated with Chronic Fatigue Syndrome (false)
link Vaccines contain heavy metals (other than mercury) (false)
link Judy Mikovits (crazy person) + this
link Didier Raoult and HCQ (crazy person)
link No saline placebo trials for vaccines (false) --- No true Scotsman fallacy

A connection between thimerosal in vaccines and autism is refuted by systematic studies but also by the simple fact that since the removal of thimerosal from childhood vaccines, rates of autism diagnosis have not decreased as predicted but actually increased.

Overview of vaccination

A month ago there was an article in Nature that summarized current efforts toward a vaccine for SARS-CoV-2.  A total of 115 different projects were identified, 73 of them clearly moving forward.  Not much specific is publicly known as yet about these different projects.

In this post, I want to start with some history about vaccination and a broad overview of vaccine strategies.

We'll take a look at specific projects for Covid in another write-up.  I'm particularly interested to read up on modern approaches and what their chances of success are.

Variolation

Smallpox has been a greatly feared viral disease that was eliminated from the earth in the 1970s.  This guy is the last person to ever have smallpox.



(Except that the Russians kept samples, and apparently that lab had a fire recently).

Both Washington and Lincoln had smallpox, Washington as a young man when he visited the West Indies with his tuberculous brother Lawrence, and Lincoln about the time of the Gettysburg address.

Variolation involves taking pus from a smallpox patient and inoculating it to an uninfected person using a needle.  Often the resulting case would be mild.  It is an old practice dating back at least to the 1500s if not before.  It was brought to England in the early 18th century.
Maitland conducted an experimental variolation of six prisoners in Newgate Prison in London. In the experiment, six condemned prisoners were variolated and later exposed to smallpox with the promise of freedom if they survived. The experiment was a success, and soon variolation was drawing attention from the royal family, who helped promote the procedure throughout England.
The problem with variolation was that sometimes the smallpox that developed was severe.  This led Jenner to his famous development of the first vaccination (from the Latin vaccus, cow).  Supposedly he noticed that milkmaids had clear complexions, completely lacking the smallpox scars that were quite common for other people.  About 1798, he took samples of a virus related to smallpox, cowpox, and used that for inoculations with great success.


Vaccines

Classical vaccines are of three types.  The first to be developed were inactivated proteins.  These were made from protein (toxins) secreted by pathogenic bacteria as they are grown in culture.  [link]  The toxins are collected from the supernatant, concentrated and inactivated with formaldehyde.
Diphtheria once was a major cause of illness and death among children. The United States recorded 206,000 cases of diphtheria in 1921, resulting in 15,520 deaths. Diphtheria death rates range from about 20% for those under age five and over age 40, to 5-10% for those aged 5-40 years. Death rates were likely higher before the 20th century. Diphtheria was the third leading cause of death in children in England and Wales in the 1930s. 
Since the introduction of effective immunization, starting in the 1920s, diphtheria rates have dropped dramatically in the United States and other countries that vaccinate widely. Between 2004 and 2008, no cases of diphtheria were recorded in the United States. 
There is a famous dog named Balto, part of a sled team that brought the diphtheria vaccine to Nome, Alaska during an outbreak in the 1920s.  I'm not sure he'd be happy about it but Balto was stuffed and currently resides in a museum in Cleveland.


[Update:  the treatment for diphtheria was not a vaccine, but an "antitoxin", that is, antibodies to the toxin protein.  Antitoxins are commonly made in large animals, especially horses.  There is a vaccine, but that came later. ]

The other types of vaccine use the actual infectious agent.  There is a vaccine for tuberculosis that is an attenuated strain of Mycobacterium tuberculosis called BCG.

A vaccine against a virus like polio may be inactivated or alternatively, a live but attenuated virus.  These vaccines depended originally on the development of mammalian cell culture and its use to produce virus in quantities large enough to immunize large numbers of people.

Inactivation is by treatment with chemicals (e.g. formaldehyde) or possibly UV.  Attenuation means that the virus grows, but does not cause disease, either because it grows too slowly, or because it is unable to infect a particular tissue (poliovirus in the nervous system).  Finding an attenuating mutant is a laborious process.  In testing pathogenicity, there is no substitute for tests with live animals.

Besides tetanus, the first group of toxoid vaccines includes diphtheria.  An example of the second is the Salk polio vaccine, and the third class includes many important vaccines including

- Yellow fever
- MMR (measles, mumps and rubella)
- Varicella (chickenpox)
- the oral (live) Sabin polio vaccine
- live (nasal spray) influenza vaccine
- Smallpox

In the last 30 years or so, vaccines have been developed for Streptococcus pneumoniae using its capsular polysaccharide, which may be chemically attached (conjugated) to a protein to make it more immunogenic.  The capsule is important in preventing engulfment (phagocytosis) by cells of the immune system called neutrophils.  Antibodies to the polysaccharide capsule allow phagocytosis and prevent disease.

Most viruses are first encountered by the host on mucosal surfaces in the mouth, the gut (digestive tract) or the airways (respiratory tract).  In these environments a particular type of antibody called IgA is predominant.

Vaccine dogma holds that to produce strong immunity, a good IgA response is essential.  The vaccine type which is best at achieving a good IgA response is thought to be an attenuated live virus vaccine, because it is the gift that keeps on giving.  Continuous presence of the antigen drives antibody producing cells (B-cells) to switch to making IgA.

Three vaccines

IPV

The news of the success of the killed polio vaccine developed by Jonas Salk was greeted with joy in 1955.  This vaccine is called the Salk vaccine or IPV (inactivated poliovirus vaccine), and it's based on three wild, virulent reference strains:  Mahoney (type 1 poliovirus), MEF-1 (type 2 poliovirus), and Saukett (type 3 poliovirus).  The three viruses are grown in tissue culture (Vero cells, a type of monkey kidney cell), and then collected and inactivated with formaldehyde (formalin).

Salk's approach was widely criticized as dangerous, and indeed, in that same year a batch from the Cutter Laboratories in Berkeley, California was not properly inactivated and it resulted in 200 cases of polio and 11 deaths.

OPV

A different vaccine using attenuated live virus was developed by Albert Sabin, called the oral polio vaccine (OPV).  It came into use around 1961.  The OPV contains the three prevalent serotypes of poliovirus which have been passaged for many generations in tissue culture and accumulated mutations that do not interfere too much with growth in the gut, but prevent neurovirulence, growth in the nervous system.  The individual components of the MMR vaccine were also developed in the 1960s.

It seems that the main reason the oral polio vaccine won out at that time was that the Salk vaccine is an injection, while the live virus is administered orally, often a drop or two on a sugar cube.  Kids prefer the sugar cube.  Also, sterile syringes are in short supply around the world.  As an attenuated live virus, the Sabin vaccine is expected to be better at inducing a good IgA response.  There is also evidence that the Sabin vaccine interferes with shedding of live poliovirus when they are both present, causing the live virus to die out.



The attenuated virus in the Sabin vaccine has been sequenced (years later).  Although there had been speculation about deletion mutations, apparently there are just a number of single nucleotide substitutions.  Only a few of these are responsible for the attenuation phenotype in any given vaccine strain.  [review]

It happens rarely that the OPV causes authentic polio, a phenomenon called VAPP (vaccine-associated paralytic polio).  The sequence of some VAPP strains shows reversion of the mutations in a one of the vaccine strains to change at least some of the attenuating mutations back to wild type.  Also, it turns out that there are other naturally occurring viruses in the same family as poliovirus (Picornaviridae, for small RNA virus), which occur in nature and these may sometimes recombine with the live Sabin virus to give a recombinant that has recovered the ability to grow well in the nervous system (ref).

In any event, this reversion is problematic not so much for the vaccine recipient (who is well on the way to immunity) but for other residents of an area with poor sanitation.  The transmission of vaccine strains to others in poor communities was originally held to be a feature, not a bug.

VAPP, which is extremely rare, finally became of crucial importance when polio was eliminated from most parts of the world.  Then, having a risk of developing authentic polio out-weighed the advantage of the oral vaccine.  When our son was born, we wrestled with the question of which vaccine to use for immunization.  We chose Salk's IPV.  Later, the vaccine schedule was changed to have the killed virus first and the live virus for boosters.

Here is a slide that shows the difference between IPV and OPV in the IgA response.


Although an IgA response is desirable, it is obviously not essential.  IPV works well.  That may well have something to do with the real pathogenesis happening in neural tissue.

Flu

The third vaccine to talk about is the live influenza vaccine.  As a virus, influenza is unusual because it has a segmented genome, each gene is on a separate piece of RNA within the virus particle.

In epidemiology, influenza is different than most viruses because it mutates rapidly enough to escape the immune response within two or three years.

As with all viruses, a major component of the immune response is directed against proteins (antigens) on the surface of the virus.  For influenza these are HA (hemagglutinin) and NA (neuraminidase).




HA (hemagglutinin) is a viral protein that binds to the receptors for influenza virus on animal cells.  These receptors are proteins with sugar chains attached to them and with the special sugar sialic acid attached at the end.  HA binds to sialic acid.  There are subtle differences in the structure of different forms of sialic acid between tissues or say, comparing human and avian hosts.  Some HA bind better to human cells.  HA is sometimes further abbreviated as H.

There is a second surface protein called NA (neuraminidase), further abbreviated N.  Its role is to cleave sialic acid from the sugar chain when newly synthesized virus is leaving the cell.  Different strains of influenza virus have different types of H and N.

The influenza virus from the 1918-20 pandemic, that has circulated in various permutations ever since, is H1N1.  The virus of the 1956-57 pandemic was H2N2, and that for 1968-69 was H2N3.

This antigenic drift means that a new vaccine must be formulated to best match the viruses that are circulating in any particular year.  The segmented genome makes it fairly easy to produce a new vaccine strain by coinfecting cells with the virus whose HA and NA you want, plus the old vaccine strain.  Reassortment of viral genome segments produces different combinations which can be screened to find the desired type that has attenuation mutations but the target HA and NA.



However, this also means that it is fairly easy (in evolutionary time) for completely new HA and NA genes to be transferred from avian flu strains to human flu strains.  This is called antigenic shift, and it happened successfully in 1918, 1956 and 1967.  The 2009 virus came from pigs (although it must be noted that we gave it to them first, in 1918, so it's not really fair to complain).

Some more slides from one of my old lectures:




Other approaches

HPV (Human papilloma virus) comes in many serotypes.  These viruses cause warts, but certain types are also associated with various cancers.

HPV is a small DNA virus and the proteins of the virus can self-assemble into a virus particle.  One HPV vaccine is made from hollow "virus-like-particles" assembled from proteins made by recombinant DNA methods, without the genome inside.

There are other approaches to making vaccines that have the promise of more rapid scale-up including DNA and RNA vaccines, and recombinant proteins (e.g. the S or Spike protein).  Also of importance are viruses that have been modified to not be harmful but can be used as carriers to express the desired cornavirus protein antigens.  These include adenovirus and retrovirus-based vaccines.

For many of these, there is a concern that they will not boost the immune system as much as is necessary to give a good response.  Various compounds (or even killed bacterial cells) act as adjuvants, treatments that increase the immune response.  Many different combinations are possible and probably most of them are being tried.

It is noteworthy that the subunit vaccine (recombinant proteins) for pertussis (whooping cough) doesn't seem to be as active as the old vaccine, which had the drawback that most kids had soreness and redness at the site of the injection for several days.

Failures

We do not yet have successful vaccines for several important targets.  These include HIV and malaria.  In the case of HIV, the success of therapeutics may have something to do with the lack of progress on a vaccine, but it is a problem where many approaches have failed.

Risks

There are a couple of negative points to mention about vaccines.

One is that it is important to keep the vaccine formulation sterile.  A tragic early event was the death of some Australian children due to injection with diphtheria vaccine that was contaminated with Straphylococcus aureus (the Bundaberg disaster).  The famous Australian scientist Macfarlane Burnet had a role in solving this mystery.

A second issue is that occasionally, for a particular vaccine, the immune response makes an infection with the virus worse.  The example we have of this is a Dengue called Dengvaxia.  However, this hasn't been seen with other viruses, or even other vaccines for Dengue.

Thimerosal

Finally, there is the issue of Thimerosal.  After problems with vaccine preparations which had been stored and allowed bacterial growth, it became the practice to add low concentrations of a compound containing mercury called thimerosal.

In the 1990s anti-vaccine activists popularized the notion that thimerosal in vaccines may have a connection to autism.  In response, thimerosal was removed from vaccines for children by 2001.  Rates of autism haven't changed, and there was never any evidence connecting thimerosal to autism.  One prominent activist, Andrew Wakefield, was found guilty of scientific fraud.  (He was motivated by payments by a group of lawyers who hoped to profit by suing vaccine manufacturers).

Non-human vaccines

There are a number of important vaccines for animals, including pets.  The "core" vaccines recommended for all dogs are rabies, distemper, parvo, and adenovirus.





Wednesday, May 6, 2020

Sensitivity, specificity and Bayes

Now that we're ramping up testing, including serology, you're going to see words like sensitivity and specificity thrown around. Here's a brief primer.

Rather than give a bunch of definitions right away, I would rather start by asking a question and then showing how we can answer it by building a table, based on three numbers (really, ratios).

I just got a positive test back for X.  Do I really have X? 

We have to recognize that there is some uncertainty here.  Peter Norvig:
suppose I have a test machine that determines whether the subject is a flying leprechaun from Mars. I'm told the test is 99% accurate. I put person after person into the machine and the result is always negative (correctly). Finally one day, I put someone (say, Tom Hanks) into the machine and the light comes on that says "Flying Leprechaun!" Would you believe the machine? Of course not: that would be ridiculous, so we conclude that we just happened to hit the 1% where the test errs. 
We find it easy to completely reject a test result when it predicts something impossible (even if the test is very accurate); now we have to train ourselves to almost completely reject a test result when it predicts something almost completely impossible (even if the test is very accurate).
I used the following example for my students:  a person walks into the doctor's office with a very bad sore throat and a rapid Strep test (for Streptococcus pyogenes) is administered.  A positive result is obtained.  What is the probability that the patient really has a strep throat?

Let's construct a table:
                        strep    No strep   total
positive test                                      pos tests
negative test                                      neg tests
total w/ or w/o strep

The question we are asking is equivalent to the following:  take the number of people with a positive test that have Strep (that's the first number in row 1) and divide by the total number of people with a positive test (the third number in row 1).

Well, how do you do that?

Sensitivity is an umpire who sees a real strike and calls "strike."  It is the proportion of people who actually have the disease that also have a positive test result.  For rapid Strep tests, this number is about 90-95%, let's call it 90.  So out of every 100 people, we write

                        strep    No strep   total
positive test              90                       pos tests
negative test              10                       neg tests
total w/ or w/o strep     100                           

Another (compound) word to describe those 10 people who have Strep but got a negative test is false-negatives.

Specificity is an umpire who sees a pitch that is actually a ball, and calls "ball".  It is the proportion of people who do not have the disease and have a negative test result.  For rapid Strep tests, this number is claimed to be about 98%.
                        strep    No strep   total
positive test              90          2            pos tests
negative test              10         98            neg tests
total w/ or w/o strep     100        100                 

Those 2 people who have Strep but got a negative test are false-positives.

Now, to get the total number of positive tests can I just add the two values in the first row?  No!!!

The reason is that the totals on the bottom need to be adjusted.  In our table so far, we have equal numbers of people that have Strep or don't have it.  

But in the real world, the actual proportion of people who walk in complaining of sore throat that have Strep is about 10%.  This is called the prevalence (or incidence if you want to be picky).  It is determined by doing further tests (streaking a throat swab onto blood agar) and other things such as PCR.

To fix this, we scale all the values in column 2 so that the total on the bottom is 900.  Then the prevalence will be right.  We do that by multiplying each value by 9.

                        strep    No strep   total
positive test              90          18     108   pos tests
negative test              10         882     892   neg tests
total w/ or w/o strep     100         900    1000         

So the probability that you actually have Strep, given a positive test, is about 90/108 = 83%.

Now, you may think this is no big deal.  83% is still a lot, even if it's not 90 or 98%.  But this is a very good test, and the prevalence is reasonably high.

Suppose we have a different example with the same sensitivity and specificity but the prevalence is 1%.

                      disease  No disease   total
positive test               9          20      29   pos tests
negative test               1         970     971   neg tests
total w/ or w/o disease    10         990    1000         

That's not so good.  Now only one-third of people who receive a positive test result actually have the disease, even with a very good test.

Sensitivity is about the ump calling strikes correctly. Specificity is about the ump calling balls correctly. Prevalence is how often the pitcher throws a strike.

Prevalence has a big influence and this is often forgotten. Since the prevalence of people who have had COVID is extremely low (first written early April, 2020), even sensitivities and specificities in the high 90s will be problematic for population-level serology testing. I leave making the table for that as an exercise...

What's the connection to Bayes?  Bayes theorem leads to a system for probability which we can think of as starting with some prior likelihood for a particular statement and then updating as evidence becomes available.

In this case our prior for the hypothesis that a patient has the disease is the prevalence.  When updated by the test result we get the final probability.

[ To do the math more easily, start with the population and first calculate disease/no disease using the prevalence.  Then use the sensitivity and specificity to get the other numbers. ]


Covid tests: LAMP + CRISPR

or, how to get to a million CoV tests a day.

LAMP

A significant limitation of PCR is the need not only to control but rapidly change the temperature of the reaction.  Normally one uses a PCR machine to do this.  These are expensive (especially with optics) and you can't just carry one around in your backpack.  People are looking for POC (point of care) tests that could be done on check-in at a local clinic, or in the home.

Twenty years ago a PCR method that works at constant temperature was invented (and patented).  It's called LAMP:  loop-mediated isothermal amplification of DNA.

The precise details are a bit too complicated to explain.  There are diagrams in this wikipedia article.



The fundamental idea is to make copies of the target DNA that contain single-stranded loops on the ends.  This allows a DNA primer for the next round of amplification to find and bind to its complementary sequence (within the loop) without the need to denature the double-stranded part of the molecule.

Another thing that's needed for this to work is a helicase activity that can open up the double-stranded DNA ahead of the polymerase in order for the new synthesis to occur, like the activity symbolized by the blue arrow in this diagram:



How that's done is proprietary, AFAIK.  People have described, for example, fusion proteins that contain both activities.

These systems can make a lot of DNA.  So much that the product can be detected by physical properties.

When a nucleotides is assembled into DNA, two phosphates joined together (called pyrophosphate) are released.  These accumulate and can be detected as a precipitate with Mg2+.   And here.

I believe this is the principle behind the Abbott POC (point of care) test for SARS-CoV-2.  The problem with Abbott's test is it requires an Abbott machine, uses expensive cartridges (vendor lock-in), and only runs a single test at a time.

We are currently running in the neighborhood of a few hundred thousand tests a day.  We need to be running millions to have a hope of controlling the pandemic since we have failed with the shutdown.

STOPCovid

"STOPCovid is an effort to develop point-of-care and at-home tests for COVID-19."

website

What they've done is to add an additional layer to the first stage isothermal LAMP process to detect the product.  The second step uses CRISPR, technology that has been in the news a lot.

To simplify a great deal, a short DNA probe is constructed for the CRISPR reaction which has an antigen A on one end and a biotin molecule B on the other.  If the probe binds a complementary sequence from step 1, then CRISPR cleaves it, separating A and B.

In the development phase, an antibody for A that is itself attached to colloidal gold is added.

With me so far?

CoV-positive:  Gold-antibody-A- plus free B
CoV-negative:  Gold-antibody-A-DNA-B

The solution flows up a slide by capillary action.

For a negative sample, the entire thing is captured on a strip of streptavidin low down on the slide, forming a visible line from the gold on the antibody.  (Streptavidin binds B, biotin).

For a positive sample, the Gold-antibody-A- doesn't stop at the streptavidin but flows further up the slide, to be captured by ____

It's cheap, does not require a PCR reaction, and could be scaled.  They are working to validate with clinical samples, but preliminary tests show 97% sensitivity and 100% specificity.  They never declare a ball to be a strike (never say someone is infected when he's not), but they do sometimes call a strike a ball.

It's about as easy to use as a pregnancy test.

[Their server is currently down.  I will update when it comes back online.]

They may have to change the name.  It's being applied to an app for tracing, used in France.





Covid tests: PCR

I'm going to write a few posts to try to make technical stuff from the news accessible to a broader audience.  Start with tests for coronavirus.  Here is a brief review of testing for SARS-CoV-2 and Covid-19 infection.

The basic test for virus is called RT-qPCR, i.e. first RT (reverse transcription) and then a modification of PCR called quantitative PCR.  We'll talk about PCR first.  Then I'll go back for RT and finally for the q.

I suppose most people know something about PCR.  PCR is short for polymerase chain reaction, we'll see the reason for that name at the end.

level 1:  basics

You start with DNA, which is usually though not always double-stranded.  If that statement is confusing, you can read up on DNA in wikipedia (and don't forget to go here to see a famous pic).  Hence, we talk about the fundamental unit of DNA as a base pair.

DNA has polarity, the two ends of each strand are chemically different.  DNA synthesis or replication proceeds with the polarity 5' to 3'.

In a diagram, the 3' end is conveniently distinguished with an arrowhead.  That's the direction of chain growth during synthesis.



In a nutshell, in PCR you may start with any old double-stranded template DNA.  It could be a known DNA like a plasmid or it could be from a clinical sample or a crime-scene specimen, or something else, as we will see.

In addition, you will need some sequence information about the template to design two short primers that have the same sequence as chosen target regions in the template.  The positions where the primer sequences are found in the template are indicated by the dashed lines in the figure above.

The primers are present in vast excess to the template, though still in single-digit nanogram amounts.

You put the template and primers in a tube with some other stuff, stick that tube in a "PCR machine" and press a button.  2-3 hours later, at the end of the PCR process, you have product, often lots of it.  The reaction product(s) can be analyzed on a gel, like this one.


level 2:

Template DNA might be anywhere from 100-200 bp (base pairs) in length up to as large as a chromosome.  As long as there are at least a hundred or so copies of template in the reaction tube, it's going to work.  It is OK if the template is just one strand, but it has to be DNA.

Primers are synthesized chemically.  When I was a student, there were specialists (chemistry graduates) who did this by hand, one base at a time.  It was a dark art.  Now they come in the mail.

The reaction requires more than just primer and template.  You need a special enzyme (DNA polymerase), plus the four nucleotides that assemble into the new DNA (technically, deoxyribonucleoside triphosphates).  Also, you need the right buffer to establish pH, set Mg2+ at the right concentration, and EDTA to chelate bad stuff.

Here's a diagram of a base (technically, a nucleotide) being added to a growing primer.

That HO group in red is a 3' hydroxyl, and it's the position where the incoming base is attached.

The key to the process is the idea of cycling the temperature.  That's Kary Mullis's innovation.

You first heat to near boiling (95°C), to denature the template (make the two strands come apart).  Then cool to the annealing temperature, somewhere between 55-60°C.  During this stage the primers bind to their locations on the template.  Then increase to 72°C and let the synthesis proceed.  (The reactions proceed under mineral oil, otherwise all the water would turn into vapor and condense on the lid of the tube).

You might think 60 C is hot, and it is.  (9/5 * 60) + 32 = 140°F, that's about the temperature of over-done roast beef or under-done chicken.  72°C is, of course, hotter.

For a DNA polymerase to function at that temperature (and not come apart at 95°C), it has to be special.  Hence Taq polymerase, from Thermus aquaticus (or various others that have been used).

So the typical procedure would be to do 30 or 40 cycles of the the 3 temperatures and then analyze what you've got.

This all happens in a machined aluminum block that accepts 0.5 mL tubes and is hooked up to heating/cooling and can change temperature quite rapidly.  The device is programmable so that running a profile is a matter of pushing two or three buttons.  They are called thermal cyclers.

level 3:

There are a couple different reactions going on, which happen with different kinetics of product accumulation.

The original template strands is copied.  The newly synthesized DNA is shown as a dashed line, but it is really just DNA.

In cycle 1, this is all that happens.  But it happens in every cycle, so this extended product accumulates linearly.  These products can be reasonably long, and end at random positions when the temperature is raised back up to 95°C at the end of the cycle, or because the template is randomly broken (sheared).

(Note:  an identical process begins with the red strand of the template, but is not shown).

In cycle 2 and beyond, the same reaction from above occurs, but another thing happens, the extended products themselves serve as templates.
Notice that this time, the chain ends precisely at the 5' end of the new template, corresponding to the sequence of primer 1.  This product occurs with more than linear kinetics, but is still rather limited.

The third reaction dominates.
                            
Here, product synthesized in a previous cycle serves as template for synthesis of another copy.  Because products serve as templates, the kinetics for this are exponential.

Assuming for simplicity that the efficiency is 100%, the amount of product doubles every cycle.  So roughly speaking, in 30 cycles you could get as much as 2^30 times as much product as template that you started with.  That's 2 to the power 30, about one billion.  Hence, the reaction is called a "chain" reaction.

This usually doesn't happen for various reasons, but you can make quite a bit of product, hundreds of nanogram.

q and RT

So briefly, the other two points.  The genome of SARS-CoV-2 is RNA.  DNA polymerases will not copy an RNA template.  However, a very well-known virus that has an RNA genome replicates to a DNA intermediate.  That virus is HIV and the family is called Retroviridae, because genetic information is flowing backward, from RNA into DNA.



Retroviruses have an enzyme packaged inside the vision called reverse transcriptase which copies the RNA genome into DNA.  Hence, the first step for RT-qPCR is a reverse transcription reaction.

The last thing is the q, for quantitative.  The original design was to use gel electrophoresis to analyze the product, like the gel shown above.  But, this is time-consuming and it's not quantitative, for technical reasons.  For diagnostics, real-time or quantitative PCR is used.

So modern labs have PCR machines with optics.  A beam of light travels through each tube as the reaction proceeds.  The reaction mixture is modified so that there is a fluorescent signal from the product, which is quantified by a detector.  It's very cool technology: there is a fluorophore and a quenching agent attached to a probe, when the probe binds the PCR product, the quencher is separated from the fluorophore and it lights up.



The usual description for a PCR reaction used in a clinical test is to talk about the Ct, the threshold cycle.  That is the cycle during which product is first detected.  (It can be determined accurately by extrapolating from data obtained in later cycles).

The smaller Ct is, the more template was present at the beginning.  A change in Ct of -1 unit means a doubling of the the amount of virus present at the beginning.



Finally, we should remember that detection of viral genome RNA does not make strong predictions about how much viable virus is present, for several reasons.  Infections with most eukaryotic viruses produce large numbers of viral particles that are not able to go on to cause another infection.  Without researching, I don't know the number for coronaviruses, but it can be as small as 1 viable virus/1000 virus particles for poliovirus (link).  Also, if viral RNA is intracellular and not packaged into a vision, it won't be infectious.

That's why when you see studies showing people testing positive for  SARS-CoV-2 viral RNA at 14-21 days it isn't anything to be concerned about.  Most authorities say that people are not infectious beyond about 5 days after they become symptomatic.

I like Carl Zimmer's writing very much, but in this article he says that "[the PCR] recipe involves many steps, which are typically carried out by trained technicians."  That's just silly.  If you have the master mix and the RT product, you mix them together, put them in the cycler and push a button.  I could teach my granddaughter Ella to do it in less time than it took you to read this.




Tuesday, May 5, 2020

Covid: good news, bad news

There's good news and bad news about Covid-19.

By simple division, using reported cases and deaths, the CFR in New York City is 10.8%.  But both cases and deaths are under-reported.  Probably, cases should be higher by a factor of 10 and deaths by a factor of 2.  That would make the CFR about 2%.

There is some hope that the CFR in the US as a whole will be somewhat less.  It's hard to say, lack of rural hospital capacity is a huge issue.  Let's call it 1%.

That's still bad news, it's about 10x the big flu epidemics in 1956-57 and 1967-68.

Covid is reasonably hard to catch.  Epidemiology studies suggest that most transmissions occur in the home, to family members or others who have extended close contact with the infected person.  Even then, with an infected person in the home, the probability of transmission is less than 20%.

However, mass transit and restaurants are places where some transmissions do occur.

The possibility of aerosol (rather than just droplet) transmission has been suggested, but this seems quite a low hazard based on estimates for the transmission ratio R of 2.5-3.  It would be a lot higher with transmission via aerosols.

I think wearing a mask indoors in public is a good idea (to prevent transmission from you to others), but a mask while walking on the beach or riding your bike it just silly.

More good news is that there is very little asymptomatic transmission, although there is significant pre-symptomatic transmission.  40-50% of transmission happen before symptoms occur.  With the mean time to onset of symptoms at 5 days, people are infectious from 1-2 days before to 5 days after.  PCR+ results from after that are not actually viable virus.

And, with aggressive lockdowns, testing and tracing, it is possible to eliminate the virus.  (New Zealand, Hong Kong, South Korea, Iceland, etc.)  Whether that's permanent or not we'll see.

However, in the US we have chosen a different path.  Both the President and conservative media have propagated the (false) idea that this is just the flu and the lockdown is unnecessary.  Add to that our natural rebelliousness.  The lockdown has only slowed but not stopped transmission.

The federal government has deliberately chosen not to develop increased testing capacity (so as to keep the apparent number of cases low), nor to develop tracing capability.  This could still help, even with if the epidemic continues.

Cases are level in the US as a whole, and some counties in the Midwest are increasing cases really rapidly.

Now we're "opening up".

I think what that means is that we really will go on to infect 50-75% or more of the country.  Estimates on the time for that range from 6 months to 2 years.

During that time I think we can expect 0.01 x 0.50 x 330M = 1.65M deaths.  That's about half the loss experienced in the 1918-20 pandemic.  I would not be surprised to be off by a factor of 2 either way.

Caveats:

- Covid may slow down in the summer, but it doesn't seem likely.

- We may develop more effective treatments.  Remdesivir early could be a great thing, but it's hard to make.

- The vaccine candidates that are least likely to be highly efficacious are the ones that will come first, perhaps in 6-8 months (nucleic acid vaccines).  But even a not-so-great vaccine could help with the CFR.