Peer-reviewedCC BY-NC-NDProceedings of the National Academy of Sciences
A semantic-based community model for high-fidelity tuning of olfactory mixture distances
23 affiliations
- 1Human Nature Lab, Yale University
- 2Department of Computer Science, Boston University
- 3Department of Statistics, University of Michigan
- 4Department of Public Health and Primary Care, Katholieke Universiteit Leuven
- 5Itec, imec Research Group at Katholieke Universiteit Leuven
- 6Department of Computer Science, University of Oxford
- 7Institut de Chimie de Nice, Université Côte d’Azur
- 8Monell Chemical Senses Center
- 9School of Electrical Engineering and Computer Science, Kungliga Tekniska högskolan Royal Institute of Technology
- 10Runway Startup, Cornell Tech, Cornell University
- 11Cold Spring Harbor Laboratory
- 12Sage Bionetworks
- 13Department of Animal Science, Texas A&M University
- 14Computer Science Department, State University of Londrina
- 15Biofisika Institutua, Consejo Superior de Investigaciones Científicas, Euskal Herriko Unibertsitatea, University of the Basque Country
- 16IKERBASQUE, Basque Foundation for Science
- 17Department of Biomedical Engineering, University of California
- 18Department of Chemical Engineering and Applied Chemistry, University of Toronto
- 19Laboratory of Neurogenetics and Behavior, The Rockefeller University
- 20HHMI, The Rockefeller University
- 21Department of Neurobiology, Weizmann Institute of Science
- 22Department of Neuroscience, University of Pennsylvania
- 23Health Care and Life Sciences, International Business Machines Research
Abstract
A central goal in sensory science is to establish quantitative mappings between physical stimuli and perceptual experience. Although such mappings are well defined in vision and audition, they remain elusive in olfaction, particularly for complex odor mixtures. Here, we show that perceptual distances between odor mixtures can be predicted with high fidelity and are unexpectedly well captured by a compact semantic space derived from single-molecule representations. In the Dialogue for Reverse Engineering Assessment and Methods Olfactory Mixtures Prediction Challenge, we assembled a unified dataset of odor-mixture pairs, benchmarked predictions on a hidden test set of 46 pairs, and integrated the top-performing models into a postchallenge ensemble. This model outperformed existing state-of-the-art approaches on the hidden test set, reducing RMSE by about 33% to 0.08 and increasing Pearson correlation by 53% to 0.57, and maintained strong performance on an independent validation set of 50 newly designed mixture pairs. An ensemble, retaining only olfactory semantic features for each model included, further improved predictions, raising the Pearson correlation by 7% to 0.61 on the test set and by 15% to 0.54 on the validation set. Given that semantic features were extracted from pure molecules, it suggests that mixture perception may not require fundamentally different representational principles from single-molecule olfaction. Together, these results establish a reproducible quantitative framework for olfactory mixture perception and advance efforts to measure, model, and engineer smell.
Related
- Mental Timing in Locomotion: Walking Across Real and Imagined Distances
- Positive early-life olfactory memory is rooted in the olfactory bulb and triggers large-scale changes beyond the olfactory system
- The role of translocator protein in olfactory mucosa neurogenesis and behavior
- Gene duplication shaped the origin and evolution of the vertebrate olfactory combinatorial code
- Review of olfactory cleft roof anatomy
BibTeX
@article{satarifard2026semanticbased,
title = {{A semantic-based community model for high-fidelity tuning of olfactory mixture distances}},
author = {Vahid Satarifard and Laura Sisson and Yikun Han and Pedro Ilídio and Matej Hladiš and Maxence Lalis and Xuebo Song and Tiffany Yang and Wenjie Yin and Aharon Ravia and Xingyu Zheng and Gaia Andreoletti and Jacob W. Albrecht and Robert Pellegrino and Wang Ze-hua and Stephen Yang and Robbe D’hondt and Achilleas Ghinis and Jasper de Boer and Felipe Kenji Nakano and Alireza Gharahighehi and José M. G. Vilar and Leonor Saiz and Benjamín Sánchez-Lengeling and Andreas Keller and Leslie B. Vosshall and Sébastien Fiorucci and Ambuj Tewari and Jérémie Topin and Celine Vens and Mårten Björkman and Danica Kragić and Noam Sobel and Nicholas A. Christakis and Joel D. Mainland and Pablo Meyer and Ruhallah Amandi and Nicola Amoroso and Loredana Bellantuono and Michele Dibattista and Alfonso Monaco and Ester Pantaleo and Sabina Tangaro and Gary Tom and Ella Miray Rajaonson and Cher Tian Ser and Sean Park and Stanley Lo and Brian K Lee and Gautam Ahuja and Siddhant Poudyal and Bableen Kaur and Douluri Pushkala Devi and Rintu Kutum and Grant D. McConachie and Soroush Arabshahi and Saeed Karimimehr and Jeremy Kotlyar and Faraz YazdaniBanafsheDaragh and Réka Böröcz},
journal = {Proceedings of the National Academy of Sciences},
year = {2026},
month = aug,
doi = {10.1073/pnas.2611057123},
url = {https://doi.org/10.1073/pnas.2611057123},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - A semantic-based community model for high-fidelity tuning of olfactory mixture distances
AU - Vahid Satarifard
AU - Laura Sisson
AU - Yikun Han
AU - Pedro Ilídio
AU - Matej Hladiš
AU - Maxence Lalis
AU - Xuebo Song
AU - Tiffany Yang
AU - Wenjie Yin
AU - Aharon Ravia
AU - Xingyu Zheng
AU - Gaia Andreoletti
AU - Jacob W. Albrecht
AU - Robert Pellegrino
AU - Wang Ze-hua
AU - Stephen Yang
AU - Robbe D’hondt
AU - Achilleas Ghinis
AU - Jasper de Boer
AU - Felipe Kenji Nakano
AU - Alireza Gharahighehi
AU - José M. G. Vilar
AU - Leonor Saiz
AU - Benjamín Sánchez-Lengeling
AU - Andreas Keller
AU - Leslie B. Vosshall
AU - Sébastien Fiorucci
AU - Ambuj Tewari
AU - Jérémie Topin
AU - Celine Vens
AU - Mårten Björkman
AU - Danica Kragić
AU - Noam Sobel
AU - Nicholas A. Christakis
AU - Joel D. Mainland
AU - Pablo Meyer
AU - Ruhallah Amandi
AU - Nicola Amoroso
AU - Loredana Bellantuono
AU - Michele Dibattista
AU - Alfonso Monaco
AU - Ester Pantaleo
AU - Sabina Tangaro
AU - Gary Tom
AU - Ella Miray Rajaonson
AU - Cher Tian Ser
AU - Sean Park
AU - Stanley Lo
AU - Brian K Lee
AU - Gautam Ahuja
AU - Siddhant Poudyal
AU - Bableen Kaur
AU - Douluri Pushkala Devi
AU - Rintu Kutum
AU - Grant D. McConachie
AU - Soroush Arabshahi
AU - Saeed Karimimehr
AU - Jeremy Kotlyar
AU - Faraz YazdaniBanafsheDaragh
AU - Réka Böröcz
PY - 2026
DA - 2026/08/04
JO - Proceedings of the National Academy of Sciences
DO - 10.1073/pnas.2611057123
UR - https://doi.org/10.1073/pnas.2611057123
AB - A central goal in sensory science is to establish quantitative mappings between physical stimuli and perceptual experience. Although such mappings are well defined in vision and audition, they remain elusive in olfaction, particularly for complex odor mixtures. Here, we show that perceptual distances between odor mixtures can be predicted with high fidelity and are unexpectedly well captured by a compact semantic space derived from single-molecule representations. In the Dialogue for Reverse Engineering Assessment and Methods Olfactory Mixtures Prediction Challenge, we assembled a unified dataset of odor-mixture pairs, benchmarked predictions on a hidden test set of 46 pairs, and integrated the top-performing models into a postchallenge ensemble. This model outperformed existing state-of-the-art approaches on the hidden test set, reducing RMSE by about 33% to 0.08 and increasing Pearson correlation by 53% to 0.57, and maintained strong performance on an independent validation set of 50 newly designed mixture pairs. An ensemble, retaining only olfactory semantic features for each model included, further improved predictions, raising the Pearson correlation by 7% to 0.61 on the test set and by 15% to 0.54 on the validation set. Given that semantic features were extracted from pure molecules, it suggests that mixture perception may not require fundamentally different representational principles from single-molecule olfaction. Together, these results establish a reproducible quantitative framework for olfactory mixture perception and advance efforts to measure, model, and engineer smell.
SN - 0027-8424
SN - 1091-6490
ER -