Thursday, November 30, 2023

The Olfactory Singularity Has Arrived

AKA Alpha Nose

Submitted to biorxiv's preprint server in September/December 2022, and published in Science September 2023, it's the first model to out-smell regular humans. If you think your sentient sovereignty is threated by a computer than can draw a picture, then it's probably time for you to get some benzodiazepines. 

You give this thing a molecule and it will tell you what it msells like. More specifically, if you type into a computer the name of a chemical, it will give you words that describe the way that chemical smells, and it will be better at doing it than a human. 

Ray Kurzweil smirks. (Because it's not 2030 yet.)

The language of smell has been a tricky thing for a long time. It became pretty obvious just how tricky when we all woke up one day to realize that you can't google smells. And then, by extension, we realized that the Internet doesn't smell, and something must be wrong, because if it's not on the internet, then it doesn't exist. 

Attempts were made to correct this. The DREAM dataset, sometimes referred to as Keller 2017, sometimes as the Rockefeller study, was the first to use the power of machine learning to crunch chemoinformatics and natural language into a prediction machine for speaking in smells. But even they had some problems, and were not able to score better than humans. Only five years later, and it's done (with the help of the Google Brain, of course).

Today, the Internet can smell.

Introductory Remarks:

  • “In olfaction, no reliable instrumental method of measuring odor perception exists, and trained human sensory panels are the gold standard for odor characterization.” (17)
  • “The model is as reliable as a human in describing odor quality: on a prospective validation set of 400 novel odorants, the model-generated odor profile more closely matched the trained panel mean than did the median panelist.”
  • The model "performs roughly on par with the median human panelist, beating a chemoinformatic baseline."
  • "The model is as reliable as a human in describing odor quality"


  • "To generate odor-relevant representations of molecules, we constructed a Message Passing Neural Network, a specific type of graph neural network, to map chemical structures to odor percepts. Each molecule is represented as a graph, with each atom described by its valence, degree, hydrogen count, hybridization, formal charge, and atomic number. Each bond is described by its degree, aromaticity, and whether it is in a ring. Unlike traditional fingerprinting techniques, which assign equal weight to all molecular fragments within a set bond radius, a GNN can optimize fragment weights for odor-specific applications."
  • "To train the model, we curated a reference dataset of approximately 5000 molecules, each described by multiple odor labels (e.g. creamy, grassy), by combining the Goodscents and Leffingwell flavor and fragrance databases."
  • Also, for novel odors, "We trained a cohort of subjects to describe their perception of odorants using the Rate-All-Tat-Apply method (RATA) and a 55-word odor lexicon."

  • called a Principal Odor Map (POM)
  • faithfully represents known perceptual hierarchies and distances
  • extends to novel odorants
  • is robust to discontinuities in structure-odor distances
  • generalizes to other olfactory tasks.

Notes of Interest:

  • The term "Odor Islands" is used when referring to certain globs of similar odors in odor space; just a cool term that was never able to exist before this model was created. 
  • Another term, "ground-truth" used while describing the model's ability to match novel odorants, "establish the ground-truth odor character for novel odorants." It's funny because the term "baseline" is corrupt in that it can sometimes refer to the previous chemoinformatics baselines, which are now inferior.
  • On Musk: "When we disaggregate performance by odor label, the model is within the distribution of human raters for al labels except musk" (which they later explain as it having 5 structural classes, as opposed to garlic or fishy which have clear structural determinants like sulfur or amines; but also the "well-documented phenomenon" of genetic variability of perception to musk.
  • On Familiarity: "[W]e see strong panelist-panel agreement for labels describing common food smells and weak agreements for labels like musk and hay."
  • On Flavor and Fragrance vs Everyday Smells: The model is better for things that have lots of training data like fruity sweet floral, less so for the less so ("ozone, sharp, fermented").
  • On Sulfur: Disaggregated by chemical class, sulfur-containing molecules showing strongest performance.
  • On Why the Language of Smell is Hard for Humans: People guess the odor wrong (aka correlation to panel mean is low) because 
1. genetic diversity for musk* (problems with the humans)
2. structural diversity like musk (problems with the chemoinformatics data)
3. unfamiliar like ozone (again problems with the humans**) 

*I thought genetic diversity was also strong for anything with a specific anosmia like putrescene or trimethylamine, then again, they didn't test "bad" smells or what I call everyday smells; the traditional Dravnieks dataset is ultimately a legacy of the flavor and fragrance industry, so it weighs heavier on good smells vs bad.

**Although unfamiliarity is a reason for this type of identification-difficulty, it should be extended beyond the individual human to our society, or maybe a bit of the fragrance industry with a bit of academia. The semantic dataset, which I will call the RATAset for "rate-all-that-apply," which is like the opposite of a multiple choice, and great for naming smells, still only uses 55 terms taken from Goodscents and Leffingwell. I would be willing to bet that more people actually know what ozone smells like, for example, they just don't have the right language at hand for naming it.  

  • On Odorant Sample Contamination: The entire section on quality control is fascinating, and news to me. "Chemical materials are impure -- a fact too often unaccounted for in olfactory research. (24: M. Paoli, D. Münch, A. Haase, E. Skoulakis, L. Turin, C. G. Galizia, Minute Impurities Contribute Significantly to Olfactory Receptor Ligand Studies: Tales from Testing the Vibration Theory. eneuro. 4, ENEURO.0070–17.2017 (2017).)"
  • Contamination, continued: Not only were there cases where the descriptions given by panelists seemingly inaccurate and later proven by GC/MS QC to be contaminated (so the panelists were right; their guess didn't match the molecule as named by the lab that sent the sample, but it did match the GCMS), but in some cases even the model got it "wrong," which implies that much of the training data is wrong, which means many of the samples of that particular chemical are likely to be contaminated. They only tested 50 of the 400 with this GCMS, but of the 50, they removed 26!
  • Contaminated Vials vs Non-Contaminated Datasets: The datasets do have words like burnt, fishy, animal, musty, sour; but these are all words that can be used to describe good parts of flavors and fragrances ("slightly burnt" or "slightly fishy"). People don't use the word semen, ever; and you will almost never see that word written in regular discourse about olfaction or the language of smell, or even when talking about linden blossoms (go right ahead, try it for yourself); it's like we're literally not allowed to talk about it. Same with the word fecal or shit or etc. There is no "dirty sock," "cigarette butt," or "cat pee" in either the Goodscents or the Leffingwell datasets. Which leads us to this --
  • They recommend characterizing the perceptual quality of contaminants.
  • "[I]t is not safe to assume that the odor percept of a purchased chemical is due to the nominal compound." (And they add that non-flavor-and-fragrance chemical commodities are not incentivized to minimize contaminants.)
  • Beyond the Perimeter of Ignorance: They created a potential odor space of 500,000 odorants "unknown to science or industry". And then then compute for us that it would take "70 person-years of continuous smelling time" to collect. (that's a lot of smelling time)
  • Limitations: The model's main limitation is that it can predict the odors of only single molecules; in the real world of perfumes and stinky trash bags, smells are almost always olfactory medleys. “Mixture perception is the next frontier,” Mayhew says. The vast number of possible combinations makes predicting mixtures exponentially more difficult, but “the first step is understanding what each molecule smells like,” Meyer Rojas says. -Scientific American Dec 2023 Machine Learning Creates a Massive Map of Smelly Molecules


via Michigan State University Department of Food Science and Human Nutrition, University of Reading Department of Food and Nutritional Sciences, Google, and Monell Chemical Senses Center: A principal odor map unifies diverse tasks in olfactory perception. Brian Lee, Emily Mayhew, Joel Mainland. Science. 2023 Sep;381(6661):999-1006. doi: 10.1126/science.ade4401.

Preprint fulltext:

Formal citation:
Lee BK, Mayhew EJ, Sanchez-Lengeling B, Wei JN, Qian WW, Little KA, Andres M, Nguyen BB, Moloy T, Yasonik J, Parker JK, Gerkin RC, Mainland JD, Wiltschko AB. A principal odor map unifies diverse tasks in olfactory perception. Science. 2023 Sep;381(6661):999-1006. doi: 10.1126/science.ade4401. Epub 2023 Aug 31. PMID: 37651511.

The Good Scents Company

Tuesday, November 28, 2023

Hope for Long Covid Parosmia Sufferers

Not one but two:

New Treatment Restores Sense of Smell in Patients with Long COVID
Nov 2023, Radiological Society of North America

Parosmia, a condition where the sense of smell no longer works correctly, is a known symptom of COVID-19. Recent research has found that up to 60% of COVID-19 patients have been affected. While most patients do recover their sense of smell over time, some patients with long COVID continue to have these symptoms for months, or even years, after infection.

The research team used a stellate ganglion block, which includes injecting anesthetic directly into the stellate ganglion on one side of the neck to stimulate the regional autonomic nervous system. The minimally invasive procedure takes less than 10 minutes, and no sedation or intravenous analgesia is necessary. CT guidance was used to position a spinal needle at the base of the neck for injection into the stellate ganglion. The researchers added a small dose of corticosteroid to the anesthetic in the pharmacologic preparation, suspecting that the COVID virus may be causing nerve inflammation. 

Follow-up was obtained for 37 patients (65%), with 22 (59%) of the 37 reporting improved symptoms at one week post-injection. No complications or adverse events were reported…

via the Radiological Society of North America and Jefferson Health in Philadelphia: New Treatment Restores Sense of Smell in Patients with Long COVID (press release). Nov 20 2023.

New clinical-trial data suggest that an antiviral pill called ensitrelvir shortens the duration of two unpleasant symptoms of COVID-19: loss of smell and taste
Nov 2023, Nature

From Japan - The medication is among the first to alleviate these effects and, unlike other COVID-19 treatments, is not reserved only for people at high risk of severe illness. The antiviral drug molnupiravir speeds recovery of these senses, but generally only the most vulnerable people can take it.

That is not true for ensitrelvir. In Japan, where it received emergency approval last year, the drug is available to  individuals with mild to moderate symptoms, regardless of their risk  factors. Its developer, Shionogi in Osaka, Japan, is continuing to conduct clinical trials of the drug, which has not yet been approved outside Japan.

In one such trial, people with mild or moderate COVID-19 symptoms were given either 125 or 250 milligrams of ensitrelvir or a placebo. At the start of  the study, 20% of participants reported some level of smell or taste loss. After the third day of treatment, the proportion of participants reporting such symptoms in the ensitrelvir groups started dropping more sharply than 
did the proportion in the placebo group.

“Most people will eventually recover on their own, but we know that some people have had 
long-term issues with smell and taste” 

via Shionogi in Osaka, and Fujita Health University: Nature. New pill helps COVID smell and taste loss fade quickly. Oct 17 2023. doi: 

Partially unrelated image credit: AI Art - The Human Condition is a Paradox - 2023

Further Reading:
Smell for Life: The Campaign to Tackle Smell and Taste Disorders
Apr 2023, Monell Center for Advancing Discovery in Taste and Smell

Thursday, September 7, 2023

Science Fiction Can Smell Too

Hella is a science fiction book by David Gerold from 2020. If you like science fiction, it's a good book by a good writer. (posting the bookshop link here in an attempt to support local bookstores?).

They take off their helmet on an alien planet for the first time. "Tell me what you smell." I sniffed. A little at first. Then a little more. "I'm not sure," I said. I inhaled again. "Something sweet. Is that grass? Something else too." I looked up at him. "Does blue have a smell?" "That's what air smells like when it doesn't come from a can." (p27)

The electronic supernose: The conical rebreather on the front of the helmet adds enough carbon dioxide to every breath so that the wearer doesn't accidentally go hyper-toxic from too much oxygen, but more important, it also sniffs the air for all kinds of particles -- it's an electronic supernose. Thehelmet integrates all this information and superimposes the augmented data onto the display. It even includes a visual presentation of all the various smells and odors and scents it can recognize. It shows us which way the scents are blowing and that helps us know from which direction any carnivores are most likely to approach. (p29-30)

About the Author - David Gerrold has been writing professionally for half a century. He created the tribbles for Star Trek and the Sleestaks for Land Of The Lost. His most famous novel is The Man Who Folded Himself. His semi-autobiographical tale of his son's adoption, The Martian Child won both the Hugo and the Nebula awards, and was the basis for the 2007 movie starring John Cusack and Amanda Peet.

Thursday, August 3, 2023

Odor Hunter Extraordinaire Cliff the Z Man Zlotnick

Cliff the Z Man Zlotnick on IAQ Radio

If you suspect a dead animal hiding in your walls (because you're house smells like a dead animal), here's what Cliff the Z Man Zlotnick does, as described back in 2007 on the IAQ Radio Show:

Put a handful of raw ground beef in a garbage bag in a wastebasket and leave it outside. Within minutes, flies will be attracted to the meat. That's when you close up the garbage bag, bring it inside, open it back up, and watch as the flies fly straight to the problem. The gases that emit from a dead animal pass through the wall itself, and the flies can smell that. 

Mandatory shout out to Avery Gilbert and his I Smell Dead People Column on his First Nerve blog (or his newer substack).

Friday, July 28, 2023

Smells Like a New Car

That new-car smell may be a sign of exposure to a host of hazardous chemicals
Apr 2023,

They tested chemicals released into the air by just one vehicle -- a brand-new, midsize, plug-in hybrid SUV in a local outdoor parking lot tested every day for 12 consecutive days using gas chromatography-mass spectroscopy.

  • Air temps ranged from 21°C to 63°C (75F - 145F)
  • 20 common volatile organic compounds tested
  • Emissions dependent on material surface temperature rather than air temp
  • Formaldehyde exceeded Chinese government safety standards at some points by up to 35%*
  • Acetaldehyde exceeded standards by 61%*
  • Benzene levels described as being unsafe for drivers breathing it for long drives
  • They suggest new car buyers ride with the windows open

via mechanical and civil engineers and occupational health scientists with several entities in China and School of Mechanical Engineering and College of Architecture and Civil Engineering at Beijing Institute of Technology, Beijing Vehicle Emissions Management Affairs Center, Beijing Products Quality Supervision and Inspection Institute, Department of Occupational and Environmental Health Sciences at the School of Public Health of Peking University, and Department of Environmental Health at Harvard T.H. Chan School of Public Health: Haimei Wang et al, Observation, prediction, and risk assessment of volatile organic compounds in a vehicle cabin environment, Cell Reports Physical Science (2023). DOI: 10.1016/j.xcrp.2023.101375

*GB/T 27630 - Guideline for Air Quality Assessment of Passenger Cars - Standardization Administration of China, Beijing - 2011

Further Reading on the Smell of the New and a sommelier describing a bunch of new cars’ smells for Car and Driver magazine back in 2003:
Baked Goods, Network Address, 2018

And for those who venture the New Jersey Turnpike:
What Exit? The Smells of the New JErsey Turnpike, Network Address, 2016

Thursday, July 13, 2023

Plume Tracking and Odor Mapping Algorithms

A deep reinforcement learning model that allows AI agents to track odor plumes
Feb 2023,

Insects track odor plumes to find mates. (And there was a similar study done recently here.)

"Instead of running a traditional laboratory wind-tunnel experiment, we used a complementary 'in-silico' approach using artificial neural networks," Singh explained. "This helped us develop an integrative understanding of plume tracking across multiple levels, including emergent behavior, neural representation and neural dynamics."

To train their plume-tracking agents using DRL, the researchers first simulated an odor emanating from a source located within a windy arena with a total area of approximately 120 m2. When their agents identified where the source of the odor was located, they received a reward. In contrast, if they lost track of the odor plume and left the arena, they were "punished."

"The behavior that emerges in our trained artificial agents bears a striking resemblance to the behavior modules biologists have previously observed in flying insects performing plume tracking," Singh said.

via University of Washington and University of Nevada: Satpreet H. Singh et al, Emergent behaviour and neural dynamics in artificial agents tracking odour plumes, Nature Machine Intelligence (2023). DOI: 10.1038/s42256-022-00599-w

Thursday, July 6, 2023

E Noses Never

Read this to learn how basically e-noses are relegated to science fiction for the next 20 years at least:

How to make electronic noses smell better
Apr 2023,

via Xi'an China Northwestern Polytechnical University: Taoping Liu et al, Review on Algorithm Design in Electronic Noses: Challenges, Status, and Trends, Intelligent Computing (2023). DOI: 10.34133/icomputing.0012

'Electronic nose' built with sustainably sourced microbial nanowires could revolutionize health monitoring
Feb 2023,

Grown by bacteria. Great, but each nanowire needs to be programmed for each molecule, so a typical top-down approach.

via University of Massachusetts Amherst: Yassir Lekbach et al, Microbial nanowires with genetically modified peptide ligands to sustainably fabricate electronic sensing devices, Biosensors and Bioelectronics (2023). DOI: 10.1016/j.bios.2023.115147

A robot able to 'smell' using a biological sensor
Jan 2023,

10,000 times higher than the usual electric-based sensors, these are now biological sensors (not sure the difference). And then they program a "library of smells", so keep in mind that, like all other smell sensors out there, these don't just smell anything that happens to be in the environment -- they can only smell things that have been pre-selected and trained-on. 

via Tel Aviv University's Sagol School of Neuroscience and School of Zoology: Shvil Neta et al, The Locust antenna as an odor discriminator, Biosensors and Bioelectronics (2022). DOI: 10.1016/j.bios.2022.114919

New devices for conveying olfactory stimuli in virtual reality
May 2023,

Aerosols and atomizers add bulk to VR gear and entail bottle filling and cleaning. This new approach uses paraffin imbued with scents, released by a temperature-sensing resistor that controls a heating element - the more heat the more scent. But wait -- magnetic induction coils pull heat away from the face to cool the wax quickly when the scent is no longer needed. 

The removal of scent is actually the harder problem to solve than the introduction of scent in these kinds of systems.

via City University of Hong Kong, Beihang University and Shandong University: Yuhang Li, Soft, miniaturized, wireless olfactory interface for virtual reality, Nature Communications (2023). DOI: 10.1038/s41467-023-37678-4