Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families
Background Amphibians are experiencing substantial declines attributed to emerging pathogens. Efforts to understand what drives patterns of pathogen prevalence and differential responses among species are challenging because numerous factors related to the host, pathogen, and their shared environmen...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
PeerJ Inc.
2025-01-01
|
| Series: | PeerJ |
| Subjects: | |
| Online Access: | https://peerj.com/articles/18901.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832575975340113920 |
|---|---|
| author | Daniele L. F. Wiley Kadie N. Omlor Ariadna S. Torres López Celina M. Eberle Anna E. Savage Matthew S. Atkinson Lisa N. Barrow |
| author_facet | Daniele L. F. Wiley Kadie N. Omlor Ariadna S. Torres López Celina M. Eberle Anna E. Savage Matthew S. Atkinson Lisa N. Barrow |
| author_sort | Daniele L. F. Wiley |
| collection | DOAJ |
| description | Background Amphibians are experiencing substantial declines attributed to emerging pathogens. Efforts to understand what drives patterns of pathogen prevalence and differential responses among species are challenging because numerous factors related to the host, pathogen, and their shared environment can influence infection dynamics. Furthermore, sampling across broad taxonomic and geographic scales to evaluate these factors poses logistical challenges, and interpreting the roles of multiple potentially correlated variables is difficult with traditional statistical approaches. In this study, we leverage frozen tissues stored in natural history collections and machine learning techniques to characterize infection dynamics of three generalist pathogens known to cause mortality in frogs. Methods We selected 12 widespread and abundant focal taxa within three ecologically distinct, co-distributed host families (Bufonidae, Hylidae, and Ranidae) and sampled them across the eastern two-thirds of the United States of America. We screened and quantified infection loads via quantitative PCR for three major pathogens: the fungal pathogen Batrachochytrium dendrobatidis (Bd), double-stranded viruses in the lineage Ranavirus (Rv), and the alveolate parasite currently referred to as Amphibian Perkinsea (Pr). We then built balanced random forests (RF) models to predict infection status and intensity based on host taxonomy, age, sex, geography, and environmental variables and to assess relative variable importance across pathogens. Lastly, we used one-way analyses to determine directional relationships and significance of identified predictors. Results We found approximately 20% of individuals were infected with at least one pathogen (231 single infections and 25 coinfections). The most prevalent pathogen across all taxonomic groups was Bd (16.9%; 95% CI [14.9–19%]), followed by Rv (4.38%; 95% CI [3.35–5.7%]) and Pr (1.06%; 95% CI [0.618–1.82%]). The highest prevalence and intensity were found in the family Ranidae, which represented 74.3% of all infections, including the majority of Rv infection points, and had significantly higher Bd intensities compared to Bufonidae and Hylidae. Host species and environmental variables related to temperature were key predictors identified in RF models, with differences in importance among pathogens and host families. For Bd and Rv, infected individuals were associated with higher latitudes and cooler, more stable temperatures, while Pr showed trends in the opposite direction. We found no significant differences between sexes, but juvenile frogs had higher Rv prevalence and Bd infection intensity compared to adults. Overall, our study highlights the use of machine learning techniques and a broad sampling strategy for identifying important factors related to infection in multi-host, multi-pathogen systems. |
| format | Article |
| id | doaj-art-0d0b6af1ea0b4d9cb962e78d14329f5f |
| institution | Kabale University |
| issn | 2167-8359 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | PeerJ Inc. |
| record_format | Article |
| series | PeerJ |
| spelling | doaj-art-0d0b6af1ea0b4d9cb962e78d14329f5f2025-01-31T15:05:09ZengPeerJ Inc.PeerJ2167-83592025-01-0113e1890110.7717/peerj.18901Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog familiesDaniele L. F. Wiley0Kadie N. Omlor1Ariadna S. Torres López2Celina M. Eberle3Anna E. Savage4Matthew S. Atkinson5Lisa N. Barrow6Museum of Southwestern Biology, Department of Biology, University of New Mexico, Albuquerque, New Mexico, United StatesMuseum of Southwestern Biology, Department of Biology, University of New Mexico, Albuquerque, New Mexico, United StatesMuseum of Southwestern Biology, Department of Biology, University of New Mexico, Albuquerque, New Mexico, United StatesMuseum of Southwestern Biology, Department of Biology, University of New Mexico, Albuquerque, New Mexico, United StatesDepartment of Biology, University of Central Florida, Orlando, Florida, United StatesDepartment of Biology, University of Central Florida, Orlando, Florida, United StatesMuseum of Southwestern Biology, Department of Biology, University of New Mexico, Albuquerque, New Mexico, United StatesBackground Amphibians are experiencing substantial declines attributed to emerging pathogens. Efforts to understand what drives patterns of pathogen prevalence and differential responses among species are challenging because numerous factors related to the host, pathogen, and their shared environment can influence infection dynamics. Furthermore, sampling across broad taxonomic and geographic scales to evaluate these factors poses logistical challenges, and interpreting the roles of multiple potentially correlated variables is difficult with traditional statistical approaches. In this study, we leverage frozen tissues stored in natural history collections and machine learning techniques to characterize infection dynamics of three generalist pathogens known to cause mortality in frogs. Methods We selected 12 widespread and abundant focal taxa within three ecologically distinct, co-distributed host families (Bufonidae, Hylidae, and Ranidae) and sampled them across the eastern two-thirds of the United States of America. We screened and quantified infection loads via quantitative PCR for three major pathogens: the fungal pathogen Batrachochytrium dendrobatidis (Bd), double-stranded viruses in the lineage Ranavirus (Rv), and the alveolate parasite currently referred to as Amphibian Perkinsea (Pr). We then built balanced random forests (RF) models to predict infection status and intensity based on host taxonomy, age, sex, geography, and environmental variables and to assess relative variable importance across pathogens. Lastly, we used one-way analyses to determine directional relationships and significance of identified predictors. Results We found approximately 20% of individuals were infected with at least one pathogen (231 single infections and 25 coinfections). The most prevalent pathogen across all taxonomic groups was Bd (16.9%; 95% CI [14.9–19%]), followed by Rv (4.38%; 95% CI [3.35–5.7%]) and Pr (1.06%; 95% CI [0.618–1.82%]). The highest prevalence and intensity were found in the family Ranidae, which represented 74.3% of all infections, including the majority of Rv infection points, and had significantly higher Bd intensities compared to Bufonidae and Hylidae. Host species and environmental variables related to temperature were key predictors identified in RF models, with differences in importance among pathogens and host families. For Bd and Rv, infected individuals were associated with higher latitudes and cooler, more stable temperatures, while Pr showed trends in the opposite direction. We found no significant differences between sexes, but juvenile frogs had higher Rv prevalence and Bd infection intensity compared to adults. Overall, our study highlights the use of machine learning techniques and a broad sampling strategy for identifying important factors related to infection in multi-host, multi-pathogen systems.https://peerj.com/articles/18901.pdfBatrachochytrium dendrobatidisRanavirusAmphibian PerkinseaRandom forestsAmphibian diseaseBufonidae |
| spellingShingle | Daniele L. F. Wiley Kadie N. Omlor Ariadna S. Torres López Celina M. Eberle Anna E. Savage Matthew S. Atkinson Lisa N. Barrow Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families PeerJ Batrachochytrium dendrobatidis Ranavirus Amphibian Perkinsea Random forests Amphibian disease Bufonidae |
| title | Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families |
| title_full | Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families |
| title_fullStr | Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families |
| title_full_unstemmed | Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families |
| title_short | Leveraging machine learning to uncover multi-pathogen infection dynamics across co-distributed frog families |
| title_sort | leveraging machine learning to uncover multi pathogen infection dynamics across co distributed frog families |
| topic | Batrachochytrium dendrobatidis Ranavirus Amphibian Perkinsea Random forests Amphibian disease Bufonidae |
| url | https://peerj.com/articles/18901.pdf |
| work_keys_str_mv | AT danielelfwiley leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT kadienomlor leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT ariadnastorreslopez leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT celinameberle leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT annaesavage leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT matthewsatkinson leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies AT lisanbarrow leveragingmachinelearningtouncovermultipathogeninfectiondynamicsacrosscodistributedfrogfamilies |