Thread by @firefoxx66

It's almost time!! Follow along with the hashtag #COVIDdynamics as we head into the first day of our three-part series on #SARSCoV2 #COVID19 Dynamics & Evolution, with this 1-day virtual conference covering the *latest* science behind the pandemic!! 🧬🙌🏽👏🏽

We start today with a session on #COVID19 Epidemiology! with @nataliexdean @JustinLessler @StephenKissler & @kathysleung covering topics like vaccine evaluation 💉, #TestTraceIsolate 🧑🏾‍🤝‍🧑🏻, geographic disparities 🌍, and the D614G mutation 🧬...

#COVID19 #SARSCoV2 #COVIDdynamics

First up is @nataliexdean - "#COVID19 Vaccine Evaluation" - reminding us the data does not exist yet! However, what should we be expecting in the data coming in the next few months?
Many trials to assure safety, immune response, & primary outcome (prevention)

#COVIDdynamics

Many vaccines in development.
RNA & DNA vaccines are designed for pandemics as are 'plug & play' - faster in theory. However - no licensed vaccines & require several doses & little evidence of efficacy.

Other methods have more history, but all uncertain at moment
#COVIDdynamics

Important message from @nataliexdean "There is more we can do in the downtime between epidemics to help our response during epidemics."
Though here this is in context of vaccines - impossible to overstate how true this is generally!!!

#COVIDdynamics

#COVID19 trails should:
- Ideally include elderly 👵🏻👴🏻(high risk)
- Randomization 💉 (vaccine vs placebo or established vaccine)
- Primary endpoints (#COVID19 disease of any severity)
- Some vaccines may prevent severe disease but not infection - still important to know!

How to make trials work?
- Smart enrolment (high risk 👵🏻, big numbers👩‍👩‍👧‍👦, hot-spots 🌡️)
- Use combination of models to predict good modelling sites 📈
- Multi-site trials as epidemic prevalence wanes & waxes across globe? 🌏🌍

#COVIDdynamics

Now we have @JustinLessler, "What makes #TestTraceIsolate work: a modelling framework & evaluation tool"
Points out it's 1 of options available with low social & economic cost - & has been used effectively in some places. But efficacy can be impacted by resources

#COVIDdynamics

Each of 3 groups (Tested, Quarantined & Community) will have different reproductive numbers (R):
Testing, if mostly at symptoms, may only prevent a few transmissions. But their contacts (Quarantine) can be caught early, pre-symptom - preventing more transmissions.

#COVIDdynamics

Of course, real transmission more complex - diffs in how asymptomatic & symptomatic tested & traced & how they transmit.

If low % of cases isolated, delays don't matter so much. As succeed in isolating higher % of cases, delays count more.

#COVIDdynamics

But really critical to remember that nobody is doing #TestTraceIsolate alone! Hard to control through this alone, but "it can make your job much easier in terms of other control measure"

In other words, it lets us get more 'normal' while containing cases!
#COVIDdynamics

And now, @StephenKissler, talking about "Geographic disparities in #SARSCoV2 prevalence are predicted by reductions in commuting-style mobility"

'Taking stock' of what happened in NYC & trying to understand variations in hospitalizations, deaths, & prevalence

#COVIDdynamics

'The story of disparities' has been a big one of the #COVID19 pandemic, says @StephenKissler.

By NYC borough, hospitalizations & deaths 2x as high in Bronx as in Manhattan - huge disparities. Could be difference in symptoms, or true prevalence.

#COVIDdynamics

Can we get semi-random testing - in a group not presenting for #COVID19 symptoms? Ex: women delivering. Showed fair positivity rate - many asymptomatic /at time/. How do we use this to estimate pop prevalence? Must consider false ➖s, as well.

#COVIDdynamics

This testing showed same disparities - much higher in Bronx & South Queens than Manhattan. Why?
Was ability to distance important? Could look at mobility data generally as proxy for distancing...

#COVIDdynamics

But this proved complicated! As mobility in borough went up - but could be walks, groceries, etc. So instead, look at community - leaving borough in morning & returning at night. And yes - commuting correlates with prevalence.

#COVIDdynamics

However, correlation doesn't imply causation! #COVID19 outcomes reflect many disparities. Need to support neighbourhoods where distancing most difficult (eg essential workers).

#COVIDdynamics

The last talk of the 1st session is @kathysleung taking a combined epidemiological & phylogenetic analysis approach to investigate the transmissibility of the much-discussed D614G mutation in #SARSCoV2! ⚡️🧬

#COVIDdynamics

D614G encodes change in S1 domain of the spike protein of #SARSCoV2. It has spread rapidly, suggesting a possible transmission advantage. It arose near end of first #COVID19 wave in China - is now majority of sequenced samples.

#COVIDdynamics

Cell culture studies show pseudo-viruses with D614G had enhanced infectivity to 293T cells expressing human ACE2 receptors - due to decreased S1 shedding & higher levels of spike protein. Can we also look with epi models?

#COVIDdynamics

Model developed to investigate relative fitness of different flu strains in 2007 & 2008 - expand to look at fitness of different #SARSCoV2 variants.
Estimates that G614 is 31% (29-33%) more transmissible than D614.

#COVIDdynamics

That estimate corresponds with COG UK estimate of 25% more transmsisible.

Does it cause more severe illness? Estimate infection fatality rate in regions with D614 or G614. Data limited, but G614 seemed to have slightly lower IFR -- though not sig different.

#COVIDdynamics

We head now into session 2 on #SARSCoV2 #COVID19 modelling!

With @C_Althaus @manonragonnet @SheetalSilal & @joshuasweitz speaking on transmission characteristics 😷, government response 🏦, near- & long-term dynamics 🗓️, & decision making ☎️.

#COVIDdynamics

We start with @C_Althaus, talking about "Transmission characteristics & infection fatality of #SARSCoV2 "

Early simulations showed outbreak trajectories in China - & emergence at end of 2019. Allowed estimates of early parameters, like R0 & doubling time

#COVIDdynamics

But R0 is just an average - also want to understand the variation in transmission! (Overdispersion, k)

What proportion of infected lead to what proportion of onward cases? Does everyone do same, or do some transmit much more?

#COVIDdynamics

This published fig shows comparison of k and R #SARSCoV2, 1918 influenza, & MERS.

#COVIDdynamics

In #Switzerland, @C_Althaus reports, 1 week of early exponential spread, required ~3 weeks of lockdown to reduce infections to the same level! 😮
"The earlier you respond, the earlier you can lift measures" - incredibly important message for #Schweiz right now!

#COVIDdynamics

Is it 'just like the flu', as 'some experts' say, asks @C_Althaus?

Estimating case/infection fatality ratio challenging:
- under-count mild cases = overestimation
- delay from illness to death means right-censoring= underestimation

#COVIDdynamics

Using a model to account for these, estimate CFR/IFR for 6 areas in Europe - IFR ranges from 0.5% (Swiss) to 1.4% (Lombardy).
These estimates are in agreement with seroprevalence studies.

#COVIDdynamics

Next up is the wonderful @manonragonnet who'll be speaking on the impact of lockdown dates on the growth of local #SARSCoV2 epidemics, as estimated using phylodynamics!

#COVIDdynamics

Lockdowns have been a big debate: do they work? Are they restriction of rights? Were they necessary?

Can we estimate the impact of lockdowns? Indeed - & some have estimated 3.1M deaths were averted in Europe. However, Sweden didn't lockdown but still had downturn

#COVIDdynamics

Can genetic data help us? Testing strategies vary across countries, making comparisons difficult. We can estimate some parameters like R0 through phylodynamic models.

Using 7 Jun GISAID data (>40,000 seqs) look at global analysis of local epidemics (31 countries)

#COVIDdynamics

Estimated Rt was correlated with mobility data from Google - found to significantly correlate. May be biased in some countries if there are associations btwn mobile phone w location & infection probability.

#COVIDdynamics

Can also estimate time of viral introduction. Many locations estimate intro in Jan/Feb - however Middle East estimates introductions later, in March.

Time from first case to lockdown varies greatly across world - & correlates w estimated # infections at lockdown

#COVIDdynamics

A big takeaway from @manonragonnet - these models ran with about 100 sequences from a location! May be a cheap and easy way to generate local estimates of #SARSCoV2 epidemic parameters.

#COVIDdynamics

Another important takeaway from @manonragonnet - people *did* modify behaviour independently of lockdown, but this is *not enough* alone to curb epidemic alone.

#COVIDdynamics

Now, we have @SheetalSilal, speaking about "the known, the unknown, and the unknowable: adaptive disease modelling to aid #COVID19 decision making in South Africa"

#COVIDdynamics

South African #COVID19 modelling consortium includes clinicians, virologists, epidemiologists & more - to help better refine parameters & assumptions to be maximally informative.

Our knowns are things we can measure - deaths, infections, etc.

#COVIDdynamics

Unknowns (for now, anyway) include things like the role of children in transmission, existing immunity, role of asymptomatic infection, & co-morbidities like TB & HIV. HIV in particular seems to have a significant increase in mortality.

#COVIDdynamics

'Unknowables' are things like R0 & heterogeneity in susceptibility

Important to remember that for many in South Africa, social distancing & quarantine are not possible if you share water source/multi-gen housing in small units. Also ordering online!

#COVIDdynamics

Modelled an optimistic & pessimistic scenario for ICU beds, deaths, cases, etc.
But as data began coming, short-term projections could be done - & models expanded w detail.

#COVIDdynamics

Our final speaker for session 2 is @joshuasweitz! He's talking about the near and long-term challenges of the dynamics of #COVID19.

#COVIDdynamics

Sorry for a short break - had to jump at the chance to advertise our public session tomorrow on @BBCNuala & @BBCOS !

Will try to pick up the thread now!

#COVIDdynamics

@joshuasweitz reminds us that large gatherings are particularly problematic:
- higher likelihood that someone has #SARSCoV2
- more potential interactions to spread (super-spreading)
- harder to contact trace - close contacts in gathering hard to reconstruct!

#COVIDdynamics

Could 'shield immunity' - large scale serological testing - reduce transmission & enable economic development?

The idea is that recovered individuals may have different interactions - this dilutes susceptible fraction to levels below herd immunity threshold.

#COVIDdynamics

In practice: could recovered health care workers preferentially treat #COVID19 patients? (Without reduction in PPE)
Models show that this 'rewiring' of interactions may have significant positive effect

#COVIDdynamics

And finally we finish with session 3 - evolution & phylogenetics - including exciting talks on genetic typing in pandemics 🗒️, & the ongoing pandemics in Brazil 💃🏻, Israel 🕍, & South Africa 🏖️ - with @AineToole @nmrfaria @Tuliodna & Danielle Miller!

#COVIDdynamics

We start the session with @AineToole! She's going to explain the famous Pangolin software (and more) along with the challenge of genetic typing for #SARSCoV2, when we have over 75,000 sequences!

#COVIDdynamics

How can we get useful info out of #SARSCoV2 phylogenies? Split the tree into lineages, clades, phylotypes.
- Phylotypes - label shared SNPs, but not persistent
- Clades are persistent, have criteria of size & SNPs
- Lineages are more fine scale - track front edge of epidemic

Lineages are defined by distinct location, evidence of local spread, & support. Size is not important - some are large, some are small. They are useful to describe many aspects of epidemic.
How to assign them as we get new sequences, though...?

#COVIDdynamics

SNPs can have issues, like homoplasies & lab issues.
Pangolin uses trees from full viral genomes - context of all diversity places tips in most likely place!
The issue is - that's a big tree!

#COVIDdynamics

Now using pangoLEARN - machine learning approach to assigning lineages! Takes all diversity into account rather than representative sequences. Overall 97.7% accuracy & 93% precision - smaller lineages have less training data and so do worse.

#COVIDdynamics

A new tool 'Llama' checks whether your sequences are monophyletic & investigate local outbreaks. Pulls out local tree from large tree from set of queries.
Similar to 'Civet' for @CovidGenomicsUK - specialised to the UK but useful for investigating outbreaks, too

#COVIDdynamics

Number one question to @AineToole? Who makes your incredible logos! 😆
Indeed - we can only all aspire to have such fantastic logos!
Turns out it's @AineToole herself! I'm sure she'll do commissions! 😉

#COVIDdynamics

Now we move to Danielle Miller (last minute switch!) to tell us about #SARSCoV2 spread into and within Israel!

Danielle starts with a reminder of #SARSCoV2 mutation rate - approx 2 mutations per genome per month - & most of these do not change the virus!

#COVIDdynamics

Looking at genetic diversity of Israeli sequences (on a local @nextstrain build!), Israeli sequences show up across the whole tree - diverse. But mostly cluster within a North America clade. Estimate 70% transmssion chains from the US - the rest mostly from Europe

#COVIDdynamics

From Ministry of Health data, 27% of returning travellers - a sig proportion of those with known location - were from USA. Did travellers returning from US contribute significantly to spread in Israel? Govt had differing policies of quarantine for US vs Europe!

#COVIDdynamics

We'll now fly back to Brazil, to hear from @nmrfaria! He'll be speaking about the evolution & epidemic spread of #SARSCoV2 in Brazil

REDCap was a successful system in Brazil, but it was discontinued in March - so reverted to old systems & a new "e-SUS VE" system.

#COVIDdynamics

Brazil has higher #SARSCoV2 incidence in more densely populated states. Little information on travel histories - early from Italy & Spain.
Basic R0 2.6-2.9 - but cases in urban areas spread to less populated locations.

#COVIDdynamics

Again, 'disparities' make an appearance: higher income & #COVID19 diagnosis - first introduced into high-income areas & spread to lower.
Difficult to asses spread in lower-income areas; challenging to get accurate diagnoses. Unequal access skews picture.

#COVIDdynamics

Brazil is 2nd highest in confirmed cases in world at 2.55 million cases 😟
After interventions, R drops from 3 to 1-1.6

Incredibly, first case went from sampling to sequence & report in only 48 hours!! An incredible achievement!

#COVIDdynamics

@nmrfaria has detected over 100 introductions, & 3 main clades in Brazil - where majority of sequences cluster.

It's possible to see the transition from within & between state transmission - international imports move to local transmission. Most are within state!

#COVIDdynamics

Finally we come to our final talk in an absolutely fantastic day - @Tuliodna will take us back across the ocean to Africa, to hear about #SARSCoV2 spread in Africa - a fitting way to finish the session!

#COVIDdynamics

@Tuliodna, sporting a fantastic new moustache 👨🏻, outlines some of his main questions: Can we estimate # introductions into Africa, and when, how, & where did community transmission happen? Which lineages? And can we use genome data to control & prevent outbreaks?

#COVIDdynamics

South Africa leads the pack for sequencing in Africa, but throughput is improving!
At the moment, can measure >300 introductions into Africa, but potentially 1000s - limited by seq data. Mostly from Europe (80%), but early intro from Asia & later from Americas.

#COVIDdynamics

Highlighting a few South Africa clusters - an introduction from Netherlands spread rapidly in a supermarket in Cape Town. Can spot a few distinct cluster expansions like this! These serve as amplification events that can fuel larger spread.

#COVIDdynamics

Like so many places, we see in South Africa that early introductions were mostly 19A (Nextstrain clades), giving way to 20A and 20B.

@Tuliodna finishes reminding that controlling the epidemic is paramount - do not want 'rolling waves' of cases!

#COVIDdynamics

And that's a wrap!! I really am totally blown away by the fantastic speakers we had today 👏🏻👏🏻
And a *huge* thank you to my amazing co-chairs @MorganeMRolland & @alison_l_hill! And to head organizer Maureen Helinski - this wouldn't have happened without her!

If you've enjoyed this thread - please join in tomorrow for our *free* public session! We'll be covering even more science & also some amazing science journalism! Please do spread the word to family, friends, & colleagues!

Register here! tinyurl.com/covidday

#COVIDdynamics