Saw the same thing first hand with Pathology data. Image analysis is far more straightforward problem than fMRI, but sorry, I do not trust your AI model that matches our pathologist’s scoring with 98.5% accuracy. Our pathologists are literally guesstimating these numbers and can vary by like 10-20% just based on the phase of the moon, whether the pathologist ate lunch yet, what slides he looked at earlier that day…that’s not even accounting for inter-pathologist variation…
Also saw this irl with a particular NGS diagnostic. This model was initially 99% accurate, P.I. smelled BS, had the grad student crunch the numbers again, 96% accurate, published it, built a company around this product —-> boom, 2 years later it was retracted because the data was a lot of amplified noise, spurious hits, overfitting.
I don’t know jack compared to the average HN contributor, but even I can smell the BS from a mile away in some of these biomedical AI models. Peer review is broken for highly-interdisciplinary research like this.
This paper gets ripped into in my favorite podcast, TWiV (This Week in Virology)!
Essentially, yes, neomycin in the nose, if timed perfectly, can activate the innate immune system, but en mass this practice would cause the spread of antibiotic resistance.
If it turned out to work great in practice and people started using it en masse, the benefits would greatly outweigh the costs IMO, particularly if it snuffed out Covid and/or the flu. If it ends up being a niche thing, I doubt it would bite into the resistance numbers.
Besides, we're already spiraling down the resistance chasm with antibacterial soaps, stuffing cattle with antibiotics, overprescribing, and so on.
Not sure what the water treatment plant steps are downstream, but I can tell you I've been processing SARS-CoV-2 positive wastewater samples for ages and managed not to catch COVID. Most of the gene fragments tested for in these studies are just that, fragments. The virus itself degrades pretty quickly under environmental stress (UV, heat, evaporation, etc).
I wouldn't be too worried about catching COVID from the water.
We're working on it! Demultiplexing the signal is tricky though.
Currently, labs have two main tools available - RT-qPCR and tiled amplicon sequencing. RT-qPCR provides cycle threshold (Ct) values which quantify the amount of virus RNA in a sample. RT-qPCR is not practical for finding and quantifying de novo variants as it relies on standard primer sets. Sequencing based approaches are less quantitative but provide more insight into the diversity of RNAs and mutations present in a sample - and thus which variants are present.
https://www.gisaid.org/hcov19-variants/ has a variant dashboard based on sequencing data, including Omicron. It's not as quantitative qPCR, but still may be of interest.
Yes! Looking for gene target failure rates is a valid, albeit indirect way of quantifying variants. You will only be able to detect variants this way if the variant contains a mutation within the primer binding region and if that mutation happens to affect primer binding causing signal drop-off (a lot of 'ifs'!).
Both of these strategies leave much to be desired. For one, it's terribly costly to run multiple RT-qPCRs in parallel. It also fails to account for any novel variants whose mutations lie outside your primer binding region(s).
From what I've gathered, RT-qPCR is useful for quantifying what is already known while sequencing helps you discover what is unknown.
I guess if you had unlimited time and money you could order new primer/probe sets from IDT every time a new variant comes into play...kind of like a home-brew microarray? Honestly, I'm kind of surprised there aren't SARS-CoV-2 variant microarrays on the market yet...we're all just spitballing here.
Folks have been testing wastewater for legal and illegal drugs for years (https://score-cost.eu has an interesting report if you're interested).
There's been talk amongst Police Departments in whether investing in wide scale wastewater monitoring could help locate drug dealers/labs. Personally, I think this is a huge invasion of privacy, but since wastewater is a "public good" police currently can do whatever they want with it (source: I work in a wastewater testing facility).
Prudence might be as big a concern. If the public gets used to reports on waste streams, it would be an easy target for those who wish to create fear and panic among the populace.
As easy as one bad actor flushing the wrong/right stuff a few times (stuff that is known to be monitored, say a specific infectious agent) in a few public toilets across a city, and waiting for the reports to trickle in. More of a fanning the flames rather than starting a panic, since the threat would have been well-known enough to have been monitored in the first place.
So maybe keep those waste reports on a need-to-know basis, rather than broadcasting a public “daily waste report”.
Also saw this irl with a particular NGS diagnostic. This model was initially 99% accurate, P.I. smelled BS, had the grad student crunch the numbers again, 96% accurate, published it, built a company around this product —-> boom, 2 years later it was retracted because the data was a lot of amplified noise, spurious hits, overfitting.
I don’t know jack compared to the average HN contributor, but even I can smell the BS from a mile away in some of these biomedical AI models. Peer review is broken for highly-interdisciplinary research like this.