Super-resolution microscopy and machine learning shed new light on fossil pollen grains

October 23, 2020

Plant biology researchers at the University of Illinois and computer scientists at the University of California Irvine have developed a new method of fossil pollen identification through the combination of super-resolution microscopy and machine learning. The team, led by Dr. Surangi Punyasena and Ms. Ingrid Romero (associate professor and graduate student in Plant Biology, respectively), developed and trained three convolutional neural network models to identify fossil pollen specimens from an unknown group of legumes.

"The global fossil pollen record is one of the most abundant terrestrial records that we have," said Punyasena. "It's the record that preserves the history of environmental and ecosystem change for the last 470 million years." The pollen record is crucial for understanding how plant species evolved and dispersed globally. By reconstructing how different ecosystems and species have changed through time, we can better understand current plant relationships and better inform conservation and climate mitigation efforts.

However, correctly measuring and identifying the morphological features of a pollen grain can be incredibly difficult. "Much of the palynological record doesn't have biological identifications associated with it," explained Punyasena. "Many of the types that we know from deeper time (beyond the last 100 thousand years or so) are groups for which we don't have a definitive sense of their identity. The effort needed to classify these types has just been too great."

Traditional methods such as scanning electron and transmission electron microscopy destroy the sample and are very labor- and time-intensive. Airyscan, by comparison, is a light microscopy method that can see below the diffraction limit of light and can be used to non-destructively collect cross-sectional images from inside and outside of the pollen grain. "This method is very useful for samples that are not abundant," said Romero. "You can mount the grains on a slide and image them efficiently without damaging the sample."

This new approach allows researchers to train machine classification models using pollen from living plants and then confirm their fossil relatives, iteratively learning from each identification to differentiate among specimens that closely resemble one another. This allowed the team to recognize genera within a larger morphological grouping of fossil legume pollen for the first time. The trained models classified fossil specimens from western Africa and northern South America dating back to the Paleocene (66-56 million years ago), Eocene (56-34 million years ago) and Miocene (23-5.3 million years ago). The most accurate model used a combination of images from both the exterior and interior of the pollen grain and was able to correctly identify samples with 90.3% accuracy.

These results suggest that Airyscan microscopy and machine learning methods could not only aid in identification of unknown specimens, but also help to constrain the time of a plant group's origin or extinction. With more information about the relationships and distribution of pollen samples in deep time, researchers can better pinpoint when and where evolutionary changes occurred.

"We have seen a huge improvement in imaging capabilities and the power of computer vision algorithms. We have gotten to the point where these algorithms can interpret complex images in a very efficient and intelligent way, giving us usable identifications. That is the key difference and innovation of these models - it's the ability to train a system on pristine modern taxa and then use that system to identify unknown and often physically distorted fossil types," said Punyasena. "This approach replicates and extends the abilities of the human analyst."
The study, Improving the taxonomy of fossil pollen using convolutional neural networks and super-resolution microscopy, is published in the Proceedings of the National Academy of Sciences.

The work was supported through funding from the National Science Foundation and the University of Illinois at Urbana-Champaign, with collaborators at Carnegie Mellon University, UC Irvine, Smithsonian Tropical Research Institute, University of Montpellier, University of New Brunswick, Missouri University of Science and Technology, and the Universidade Federal de Mato Grosso.

University of Illinois College of Liberal Arts & Sciences

Related Pollen Articles from Brightsurf:

Modern computational tools may open a new era for fossil pollen research
By integrating machine-learning technology with high-resolution imaging, scientists are improving the taxonomic resolution of fossil pollen identifications and greatly enhancing the use of pollen data in ecological and evolutionary research.

Efficient pollen identification
From pollen forecasting, honey analysis and climate-related changes in plant-pollinator interactions, analysing pollen plays an important role in many areas of research.

Scientists shed new light on pollen tube growth in plants
New insight on how an enzyme ensures the correct growth of pollen tubes in flowering plants has been published today in the open-access journal eLife.

Fossil pollen record suggests vulnerability to mass extinction ahead
Reduced resilience of plant biomes in North America could be setting the stage for the kind of mass extinctions not seen since the retreat of glaciers and arrival of humans about 13,000 years ago, cautions a new study published August 20 in the journal Global Change Biology.

Researchers explore pollen fertilization mechanisms
A study showing how pollen tubes grow into flowers to reach the ovule paves the way for the improvement of food crop varieties as well as a deeper understanding of the growth of fungi and neurons.

The mystery of pollen sterility and its reversion in pigeon pea revealed in a new study
The Vienna Metabolomics Centre (VIME), University of Vienna, in collaboration with International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), based in India has made a breakthrough in pigeonpea by resolving the mystery behind fertility-sterility transition in pigeonpea.

Newly identified gene reduces pollen number of plants
Producing less sperm cells can be advantageous in self-fertilizing plants.

'Bee' thankful for the evolution of pollen
Over 80% of the world's flowering plants must reproduce in order to produce new flowers, according to the US Forest Service.

Spring rains are a surprising source of pollen
Spring rains washes away some pollen, but not all. University of Iowa researchers have found tree pollen fragments can remain airborne for hours after a storm.

Nasal biomarkers predict severity of pollen-specific allergy symptoms
It is not only people with allergies, but also a subgroup of people without one that suffer in spring and have nasal problems from contact with pollen.

Read More: Pollen News and Pollen Current Events is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to