India's new paper Covid-19 test | Hacker News Reader

India's new paper Covid-19 test

bbc.com

169 pointshardmaru28 comments

The preprint Rapid, field-deployable nucleobase detection and identification using FnCas9 [1]:

> Here, we present FnCas9 Editor Linked Uniform Detection Assay (FELUDA)... At the heart of such a detection procedure lies the property of CRISPR proteins to accurately bind to target DNA or RNA, undergo conformational changes leading to cleavage of targets generating a reporter-based signal outcome.

> We then developed a pipeline to adapt FELUDA for an affinity-based fluorescent read-out system, where the amplification step generates biotinylated products which can then be immobilized on magnetic streptavidin beads.

The end-to-end process is similar to PCR and requires specialists to collect samples that are then processed by a lab.

[1] https://www.biorxiv.org/content/10.1101/2020.04.07.028167v2

Interesting, I was led to believe this kind of research could only happen when we let drug companies gouge us for obscene profits.

Right? Very good job from CSIR Govt. Labs - https://www.csir.res.in/

CSIR is more of a system of labs than one big organization/lab. It provides a legal framework for the state to start targeted labs which investigate specific fields of study. There are CSIR labs studying everything from ceramics to biology.

About the name: Feluda is a famous detective character [1] created by the director Satyajit Ray [2]:

[1] https://en.wikipedia.org/wiki/Feluda

[2] https://en.wikipedia.org/wiki/Satyajit_Ray

Michael Mina [1] has promoted [2] the idea of $1 antigen tests as a screening tool. The idea is not that these are diagnostic (medical) tools, rather they are tools to use in schools, workplaces, sporting events. A positive antigen test would be followed up by a PCR test. The "This week in Virology" hosts had Mina on their podcast [3] and fully buy this argument for cheap tests for screening

It seems to me that this test from India is going in the same direction, though it's not clear why it's so much more expensive than what Mina describes.

[1] https://twitter.com/michaelmina_lab

[2] https://www.rapidtests.org/

[3] https://www.microbe.tv/twiv/twiv-640/

> It seems to me that this test from India is going in the same direction [as the antigen rapid tests]

I think the FELUDA CRISPR/Cas9 test is more of an alternative to PCR with the potential for significantly reduced price per test. Michael Mina's $1 Lick-a-Stick self-administered rapid test is much more of a game changer. The BBC article said that "...the new test had 96% sensitivity and 98% specificity." The fact that we are talking about sensitivity without specifying the viral concentration of the test sample means that Mina's message is not getting through.

We need rapid tests that target infectious viral loads, that is, equivalent to PCR Cycle Threshold (CT) of 24 or less.

EDIT: changed $5 Lick-a-Stick to $1 - the Abbott BinaxNOW test is $5

According to the article, this is based on "Crispr gene-editing" technologies. I am no gene-specialist, but apparently it "latches on to a set of letters of a gene carrying the signature of the novel coronavirus, highlights it [with] two blue lines [to] indicate a positive result".

I feel like none of that sounds like it's got much to do with gene editing? CRISPR itself is just a system of DNA, right, but in itself has not got much to do with gene editing?

CRISPR is more commonly paired with Cas9. The CRISPR part is the bit that latches onto a specific DNA sequence, and the Cas9 part cuts the DNA at that particular location. So CRISPR is key to gene editing, but can be used in other things like this.

$6.75 seems pretty expensive for a "ground-breaking" low cost test kit. What's the scaled cost of current tests without profit and overhead?

I don't know how much it costs in India, but I have paid around 75 Euros in France, and I think it's around 50 Euros in Germany. All of these prices which I cited for France/Germany are for those who don't have any health insurance. So reduction of cost by 8-10 times should be commended.

But for a poor Indian, 7$ is still a substantial amount of money. For a middle class Indian, it's quite affordable though.

TL;DR:

Cost: $7

Results: 1 hour

Sensitivity: 96%

Specificity: 98%

If these numbers hold up, this could be a game changer.

I hope this is real and not an Indian Theranos

Please !!! Read the article; This is from CSIR Govt. Labs and not some BS company CEO.

I have already witnessed false-positives with much more elaborate Covid tests so I would be surprised if this test is actually realiable.

Did you read the text before making this comment?

> They found that the new test had 96% sensitivity and 98% specificity.

I'm not sure if thats too good - 98% specificity means that for 1000 infected people, 20 will get away unidentified (false negative). But probably better than other tests, i guess?

But I guess that's why, in some countries, people who had a high probability of catching the virus are told to stay in quarantine even if they tested negative.

The text states this is comparable to PCR tests (molecular tests). So it is really not bad.

In the UK, only 2.7% of tests are returning positive results. At the end of July, this figure was 0.4%, which puts a floor on the specificity of the existing testing technology at 99.6%.

If a test with only 98% specificity were used in the UK, the test would have a positive predictive value (the chance that the positive test represents a true positive result) of about 50%, far worse than existing technology.

Yeah, that's the fundamental problem with this kind of cheap, quick test - it's really hard to get good enough specificity to make them viable. This is part of the reason the UK has rejected pretty much every such test it's tried, with Matt Hancock infamously declaring that "no test is better than a bad test" (though I think those tests were rather worse than this one).

WHere are you getting 2.7% from. In the last 24 hours in Scotland 12.8% of people tested were positive.

It's been above 10%, I think, now for over two weeks.

UK wise there's about 10,000 positive cases a day from 250,000 PCR tests

E.g.

01-10-2020: 255,915 PCR tests processed, 11,988 cases dated then

Obviously the testing cockup hasn't helped, but we're still on the order of 4% of tests coming back positive

Scotland specifically hasn't had the cockup, Oct 1st was 762 cases (668 reported) from 14,316 cases, or about 5%.

See here:

https://www.gov.scot/publications/coronavirus-covid-19-daily...

697 new cases of COVID-19 reported; this is 12.8% of newly tested individuals. Are some people tested multiple times and not a their tests come back positive?

That's an odd thing to report

"The number of people newly tested refers to the number of individuals in Scotland who have been tested for the first time for COVID-19 on the previous day. Note that each person is only counted once regardless of repeat tests. This means that where someone tests positive after receiving a negative test result on a previous day, they would be counted in the new positive cases for that day but in the number of people newly tested on the day they first received a test result."

So if I were tested in August as I had a cough, and it came back negative, but now I have a temperature, and test positive, I wouldn't count as a newly tested person.

The 697 new cases would include my number, but the 5445 newly tested people wouldn't.

It's very possible for new cases to outnumber newly tested people, so dividing one number by another makes no sense at all.

It's concerning that the Scotish government are linking these two numbers.

https://ourworldindata.org/coronavirus-data-explorer?zoomToS...

Where are you getting 10% from???

Today's numbers from the government. They release daily.

https://www.gov.scot/publications/coronavirus-covid-19-daily...

> 697 new cases of COVID-19 reported; this is 12.8% of newly tested individuals

Is There a difference between number of people tested and number of tests carried out? Ie, do some get multiple tests?

Reply on news.ycombinator.com