We are looking to recruit a Data Scientist.

Job Description
The Signature Research Programme in Neuroscience and Behavioural Disorders (NBD) is focused on understanding the structure and function of the nervous system, and the neural mechanisms underlying human neurological, psychiatric and ophthalmological disorders.
The Claridge-Chang Laboratory, Duke-NUS Medical School is looking for a data scientist to help transform the way scientists analyze data. The position is funded for two years. Recently, the laboratory has led the way in introducing a new way of data analysis and estimation statistics. It has achieved this progress with a suite of software, Data Analysis with Bootstrapped Estimation (DABEST), available as libraries in R, Python, and a web application. These software tools have wide usage, representing an important shift in statistical culture in biomedical research.

• https://github.com/ACCLAB/DABEST-python
• https://github.com/ACCLAB/dabestr
• https://www.estimationstats.com/#/

The existing software covers five of the most widely used analysis models, suitable for a variety of small data sets and common experimental designs. The laboratory has won a competitive grant and the candidate will work on DABEST for a wider set of data types, including visualizing large data with estimation graphics. This project will be in collaboration with Assoc Prof. Hyungwon Choi, NUS Medicine.  The candidate will benefit from the dynamic environment of two laboratories with world-class expertise in experimental design, advanced statistical analysis of high-dimensional data, and data visualization.
The role will involve both software development for new modules of estimation statistics, and application of estimation methods to exciting research questions, encouraging a productive dialogue between analysis methods development and application. The selected candidate will perform a variety of research activities, including but not limited to the following:

• Develop the DABEST visualization package for new categories of estimation data analysis.
• Develop an empirical Bayes method for analysis of high-dimensional analysis and visualisation with estimation.
• Document the code and write introductory materials to introduce the new methods.
• Record instructional videos introducing the methods and describe the new methods to a broad audience.
• Coordinate the estimationstats.com web application.
• Analyse and process data from neuroscience and proteomic experiments, using estimation and other statistical methods.
• Contribute to project management, presentations and publications of research work
• Provide guidance to junior researchers as well as undergraduate and graduate students.
  

Job Requirements

• PhD or Master’s Degree in Statistics, Data Science, Bioinformatics, Computational Biology or related areas with demonstrated expertise in statistical computing and data visualisation.
• Experience in programming and software development, as well as machine learning would be advantageous.
• Experience with statistical modelling and analysis of large-scale data.
• Has good programming skills in R and Python.
• Demonstrate good knowledge, skills and expertise, with potential to achieve research and methods-development excellence.
• Able to effectively multi-task and perform well under pressure, both independently and as part of a team.
• Able to prioritise and work collaboratively in a diverse workforce.

ApplyPlease email a CV and cover letter to claridge-chang.adam<at>duke-nus.edu.sg.

This article was originally published in the Straits Times 10 February 2022

On my way to work, I often walk past a smoking area. While second-hand smoke can harm passersby, cigarette smoke can at least be seen and smelled so the risk to others is apparent. Like tobacco smoke, the Covid-19 virus is also airborne: it is exhaled, floats around and can fill a room. But unlike smoke,  its airborne particles are odourless and invisible.

The transmission mode of Covid-19 has been controversial. Some people still believe that the main routes of transmission are through touching contaminated objects that can carry infection such as doorknobs, or through coming into contact with mouth droplets.
However, a large number of studies have now demonstrated that the major route of transmission is via tiny airborne particles called aerosols.

So the announcement a week ago by Singapore’s Building and Construction Authority (BCA) that it is launching a major study into the use of natural ventilation to reduce transmission inside buildings is welcome and timely (The Straits Times, Jan 30, ‘Study to look at reducing disease spread in Singapore buildings’).

As the planned study suggests, a simple, effective approach to deal with airborne transmission will be to open buildings to allow more outdoor air.
Innovating modern versions of traditional fan-based cooling will make indoor spaces comfortable, while facilitating the elimination of contaminated air. In the short term, ventilation can be improved by opening windows, and holding gatherings outdoors whenever possible.

However, not every indoor space can be opened up, and air conditioning will remain a reality. How can that issue be addressed? And how did it take so long to move away from the focus on droplets, and recognise Covid’s airborne transmission?

Droplets v aerosols
The critical difference between droplets and aerosols is that droplets are so large that they fall to the ground within a few metres in seconds, while aerosol particles are small enough to float in the air for minutes or even hours.

In March 2021, a group of scientists led by Trisha Greenhalgh, professor of Primary Care Health Sciences at Oxford University, published a short letter listing 10 points of evidence supporting the Covid virus’s airborne transmission. One devastating observation was the frequency of long-range infections.

One such outbreak occurred in a restaurant in Guangdong. At three adjacent tables, nine people were infected by the index patient. Security video confirmed that the index patient never had close contact with the other tables - so droplet transmission was ruled out. All three tables were cooled by the same air conditioner. Some of the infected people were right next to the airconditioner’s blower, and so were seemingly “upwind” of the index patient.

A team of scientists, headed by Min Kang at the Guangdong Provincial Centre for Disease Control and Prevention, used smoke to track airflow. They discovered that the air was recirculating, and that the “upwind” infected people were effectively 10 metres downwind of the index patient.

Long-distance outbreaks are common - hence the word “super-spreader”.
Early in the pandemic, on February 11 2020, the World Health Organisation’s director-general, Tedros Ghebreyesus, got it right: he announced  the virus’s airborne transmission.

He said: “This is airborne. Corona is airborne.” But in the following month, WHO officials backtracked. Dr Tedros himself wrote “ ... actually it’s not airborne”.

But throughout the first year of the pandemic, evidence for airborne transmission accumulated and scientists increased the pressure, culminating in Professor Greenhalgh’s letter. On April 30, 2021, rather quietly, WHO updated its website to acknowledge that aerosol transmission plays a role. Even today WHO appears to minimise airborne transmission.

WHO’s stance can influence national policies. Currently, the Singapore Ministry of Health’s (MOH) main website, under the heading “How does Covid-19 spread?” starts off by echoing WHO’s old droplet claims.

It adds later on: “While there have been limited reports of airborne transmission outside of health care settings internationally, its role and extent are under further study. MOH will continue to monitor the evidence as it emerges.”

Nevertheless, since 2020, MOH, together with the National Environmental Agency, and the Building and Construction Authority (BCA), has been advising ways to improve ventilation to fight Sars2 (as the Covid-19 virus is known). And last year, the agencies advised that Sars2 “can also be spread through virus aerosols in the air under certain settings, such as enclosed environments which are poorly ventilated”, along with updated advice to building owners and facility managers on how to improve indoor air quality. Indeed, the adoption of some of these airborne-oriented measures might have already helped to reduce transmission.

What can and needs to be done
And barely a week ago came the BCA’s announcement looking into natural ventilation.
Natural ventilation is an excellent solution for many settings, however not all indoor spaces can be opened up in this way, and air conditioning will continue to play a role. So we also need to invest in better air monitoring, filtration, and purification to keep our indoor air clean.

Carbon dioxide monitors can measure how much air in a space has already been breathed. Particulate monitors can measure the extent of particle and aerosol contamination.

Air purifiers with HEPA-type (High Efficiency Particulate Air) filters clear aerosols effectively. As an interim solution, affordable do-it-yourself purifiers can be constructed and installed within a day, before buildings’ air-handling systems are upgraded. If HEPA-type filters had been used in the Guangdong restaurant’s air conditioner, that outbreak might have been prevented.

Ultimately, we need new building codes and designs that will make indoor air as safe and healthy as outside air. 

David Fisman, professor of Epidemiology at the University of Toronto and co-author of the Greenhalgh letter jokes, “Air: it’s the new poop.” People in wealthy countries don’t tolerate faeces in their water supply because it causes potentially fatal diseases. So why should they tolerate workplaces, schools, or restaurants with potentially deadly exhaled air?

Or an even closer analogy: I’m old enough to remember when indoor smoking was allowed, and a bar could contain so much exhaled smoke it would be hard to see across the room. Now it seems absurd that we would allow public spaces with high levels of visible carcinogenic aerosols.

One day, we will think the same way about invisible virus aerosols. To control the coming waves of  Covid variants - and the next pandemic - let’s hope that day comes soon.

-Adam Claridge-Chang

By Adam Claridge-Chang.
A version of this opinion piece was first published by the South China Morning Post on 14 October 2021.

Singapore’s vaccination rate is one of the highest in the world, above 80 per cent. Encouraged by this, the city-state has tried to reopen and renormalize. However, the recent COVID-19 surge has grown faster than expected, hospitals are feeling the impact, restrictions were reintroduced, and mobility remains well below pre-pandemic levels.

Here and abroad, there is a vigorous debate between two groups on how to ‘exit’ from the pandemic.

On one side are the ‘eliminationists’. They point to China which advocates that the safest way to defend against COVID-19 is to eliminate it. Lockdowns and quarantines, they argue, are a price worth paying to keep everyone safe. They highlight that eliminationist economies have overall grown faster than mitigation economies. Eliminationists note the potential threat of novel variants and that long COVID is incompletely understood.

On the other side are those advocating to re-open society. They argue that, since vaccines protect the vast majority from severe disease, allowing Delta to spread is acceptable and even desirable for post-infection immunity. The re-openers point to well-vaccinated countries in Europe that lifted regulations without major spikes in deaths. Re-openers say that ongoing travel and social restrictions will do needless damage to society, mental health, and the economy. 

Both camps make strong cases. But what if we could do both? What if we could strongly suppress COVID-19 while also re-normalising society? Some believe this can be done with at-home testing.

In the pandemic’s first days, the WHO urged the world to “test, test, test”. By identifying and isolating cases, they said, the chains of transmission could be broken. Polymerase chain reaction (PCR) testing—a powerful way to detect viral genes (RNA)—was widely implemented. 

Later, another kind of test became available: the antigen rapid test (ART). ART kits are like a pregnancy test for COVID: they require no special electronics or skills, and can be used by anyone. For the past year, Dr Michael Mina, an assistant professor of epidemiology at the Harvard T.H. Chan School of Public Health in the US, has been advocating that the widespread use of ARTs could force COVID-19 transmission into collapse. He bases this on ART’s three critical advantages.

First, ARTs can reliably detect if someone is in the roughly weeklong infectious phase, when they secrete millions of viruses. This high-virus phase corresponds to the interval when ARTs are highly sensitive, making them an almost ideal tool for someone to know when they are contagious. Critics of ARTs say that PCR is much more sensitive, but this can actually be a liability when it detects a residual amount of RNA during a waning, non-infectious case. Data from a study by the UK Covid-19 Lateral Flow Oversight Team indicates that ART sensitivity for contagious cases, even when asymptomatic, is typically around 97 per cent. This means that, with rapid testing, positive individuals can reliably know which few days they need to stay home, while everyone else testing negative can confidently get on with their lives. 

Second, ARTs are very fast. The Delta variant becomes infectious 1–2 days earlier than last year’s strains, effectively outrunning PCR tests, which can take days to give a result. An ART can let someone know if they are contagious in just 20 minutes, beating Delta’s pace.

Third, ART technology lends itself to mass production at low cost (although in many places this has not yet been fully realized). ARTs will be most effective when testing is affordable for everyone. Transmission will be more effectively slowed as ARTs are used more widely. Therefore, better control will be achieved when ARTs are low cost or, ideally, free.

Indeed, some European nations have made low-cost ARTs integral to their re-opening strategy. In Germany, the federal regulator has authorized dozens of different ART brands, driving prices under €2.00. In the UK, households can order a free pack of seven tests every day, or collect free test packs from local collection points. 

However, places like the United States and Singapore have been slower to increase ART availability. The US has stringently regulated ARTs as medical devices, resulting in only a handful of approved home-test vendors, high prices, and shortages. President Biden is moving to fix this, invoking the US Defense Production Act to help increase production and directly buying millions of tests.

In Singapore, one especially clever application is using ARTs in coordination with automated contact tracing: close contacts are given free kits and expected to test daily for a week in case they become positive. However, approvals of home-use ARTs have been relatively slow—just a few brands approved at the time of this writing. Prices at the major retailers remain high, around S$10 per test. In Germany, an ART costs roughly the same as a ride on Singapore’s Mass Rapid Transit (MRT). Broadening vendor approvals should lower prices, so that everyone can afford to test regularly. A survey in the US showed that, if available for U$1.00 (S$1.35), as many as 79% would use ARTs regularly. Low cost would also make it easier for governments and employers to distribute kits freely to citizens and staff.

Whatever we do now is preparation for the next variant wave or future pandemics. Self-testing is already as easy as other hygiene routines like tooth-brushing. We know it can be made as cheap as a daily commute. When that happens, we will find out whether we can crush the pandemic with a new kind of ‘MRT’: mass rapid testing.

A few notes on typography for science communications. For an excellent course in typography, read Butterick's Practical Typography. Butterick's book is useful to learn general typography, especially this short list.

In specific terms, many scientists work in Google Docs and slideshows. To set type in these apps, I recommend some professional-quality typefaces that are (1) free with an open SIL license, (2) available both in GDocs and by download for use in desktop apps, and (3) has Greek letters and other symbols needed for biomedical research communications. Some of the best of these are:

• IBM Plex Sans. Maybe the most versatile, works well for headings, text body, slides. Has a lot of weights to separate elements. There is a mono-spaced version for tables of numbers. 
• Fira Sans. Also very versatile for all purposes. Great in slides, it has a bit more flair than Plex. Also a lot of weights, and a Mono version.
• Source Serif Pro. As a serif typeface Source Serif is a strong option for text body. I think it looks nice with Fira for headings in a document. (In case you sometimes write documents in Chinese, Source Serif even has a CJK version called Source Han Serif.)
• Inter.  Especially good for labeling Figures, since it has that neutral, generic interface-font appearance. However, this strength makes it weaker for text body. Could also be paired for headings, eg in combination with Source Serif.

If you don't use GDocs, two other options are STIX and Brill. Both can be downloaded and used freely for non-commercial use. Their licenses don't appear as permissive as the four above, which probably explains why Google hasn't grabbed them for GDocs.

• STIX is a workhorse typeface commissioned by big players in the science publishing industry, it has pretty much every special scientific symbol you could ever possibly need. This would be a great option to set mathematics, as it has symbolic libraries specifically designed for that purpose.
• Brill was designed for an ancient academic publishing house (also called Brill), and also has a huge glyph set. I like that it has an old-school look, would be ideal for typesetting a doctoral thesis.

My old workflow was to write in GDocs and then download to Papers to typeset. But since GDocs added Fira Sans, Plex, and Source Serif got proper italics, I now only rarely see a reason to do this and typically export from the GDoc to a PDF directly.

Before the days of corporate-sponsored high-quality open-license free typefaces, these would have each cost about $500 to license. 

One of the services provided by journal publishers is to typeset a paper nicely, to make it attractive and readable. However, with the rise of preprints and publishing to the web, publishing is changing, and more scientists need or want to typeset their own writing. Typography will improve as we move away from PDFs and start to embrace programmatic typesetting, such as eLife's Lens software, which presents an xml file in a side-by-side format with hyperlinks between Figure tags in the body and the Figures themselves. 

In 2018, we wrote a short editorial to introduce new users to estimation graphics. We first released this as a preprint, which was received well.

We then submitted this short piece, with the software for consideration by Nature Methods. After some time, and addition of several features, this was accepted and published. The publication was also met with a warm reception by scientists.

Several other statistics-reform advocates recommended the software for scientists to use in their data analysis. Lewis Halsey, who previously highlighted the volatility of p-values, noted that DABEST and the estimationstats web app are the best methods to produce publication-quality estimation graphics.

Most recently, the Editor-in-Chief of eNeuro, the Society for Neuroscience's open-access journal, announced that he was changing the journal's policy to encourage the use of estimation statistics. This was backed up with a great explanation of the why and how of estimation by Robert J. Calin-Jageman and Geoff Cumming. The Editor-in-Chief Christophe Bernard wrote a personal account of how he came to adopt estimation:

This is a very promising development. We hope that this policy change at eNeuro spreads further, as we believe journal policy, when paired with the right tools for analysts, is a powerful way to transform statistical practice.

In November 2018, we celebrated Farhan's new job as an Assistant Professor at Hamad Bin Khalifa University in Qatar. While we were sad to see him leave, we are thrilled that he will be starting the next phase in his career. He is more than ready to take on the challenges of teaching and doing independent research.

On the 19th, 20th and 27th of May 2018, members of the lab volunteered to help introduce Drosophila research to children at Singapore's Science Centre.

James, Joses, Sangyu, and Quyen spent three weekend afternoons interacting with kids showing them Drosophila ​flies and experiments in person. Hundreds of children had fun looking at flies!

Four decades of memory genetics
We systematically reviewed over 30 years of Drosophila memory genetics. While replication is sparse, the replicated genes have high reproducibility. There is currently almost no evidence connecting genes into plasticity pathways.