Lara Marie Kösters
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Lara Marie Kösters has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Ladislav Hodač
Biodiversity, Ecosystems and Society Group, Max Planck Institute for Biogeochemistry
4 shared publications - Jana Wäldchen
Max Planck Institute for Biogeochemistry, Jena, Germany
3 shared publications - Kevin Karbstein
Biogeochemical Integration, Max-Planck-Institute for Biogeochemistry , 07745 Jena
3 shared publications - Martin Hofmann
Data-Intensive Systems and Visualization Group (dAI.SY), Technische Universität Ilmenau, Ilmenau, Germany
3 shared publications - Patrick Mäder
Technische Universität Ilmenau
3 shared publications - Susann Wicke
Institute for Evolution and Biodiversity, University of Muenster, Muenster, Germany
3 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (9) , e1014722
Balanced DNA interpolation improves learning of genetic distance-informed embeddings in plants
In taxonomic research, traditional phylogenetic tree and structure analyses of genetic data are increasingly complemented by machine-learning-based identification and representation learning. Although the amount of DNA data needed to train state-of-the-art machine learning models often exceeds what can realistically be collected and sequenced in biological studies, the number of samples can be extended artificially through data augmentation. Genetic data augmentation usually refers to the introduction of random base variations, translocations, and reverse complementing. These augmentations do not take into account the inherent structures of populations and species, potentially blurring the lines between entities within genetic datasets. Here, we propose DNAInterpolator, an approach based on interpolation of DNA sequences within a given dataset that presents a neighbor-guided alternative to random mutations. We tested interpolation as an augmentation technique using four flowering plant datasets and an artificial neural network trained to predict genetic distances between paired samples. To address unequally distributed distances within our training datasets, we examined the effect of balancing the distance distribution by curating interpolated sequences. We found that balancing helps models capture genetic distances across the full distance range by strengthening performance in underrepresented regions of the distribution. Our new approach leverages the potential of taxonomic DNA datasets for modern machine learning applications.
- 3 September 2026
- Published by Public Library of Science (PLoS) in PLOS Computational Biology
- Vol. 22 (9) , e1014722
Open AccessBalanced DNA interpolation improves learning of genetic distance-informed embeddings in plants
In taxonomic research, traditional phylogenetic tree and structure analyses of genetic data are increasingly complemented by machine-learning-based identification and representation learning. Although the amount of DNA data needed to train state-of-the-art machine learning models often exceeds what can realistically be collected and sequenced in biological studies, the number of samples can be extended artificially through data augmentation. Genetic data augmentation usually refers to the introduction of random base variations, translocations, and reverse complementing. These augmentations do not take into account the inherent structures of populations and species, potentially blurring the lines between entities within genetic datasets. Here, we propose DNAInterpolator, an approach based on interpolation of DNA sequences within a given dataset that presents a neighbor-guided alternative to random mutations. We tested interpolation as an augmentation technique using four flowering plant datasets and an artificial neural network trained to predict genetic distances between paired samples. To address unequally distributed distances within our training datasets, we examined the effect of balancing the distance distribution by curating interpolated sequences. We found that balancing helps models capture genetic distances across the full distance range by strengthening performance in underrepresented regions of the distribution. Our new approach leverages the potential of taxonomic DNA datasets for modern machine learning applications.
- Open Access
- 13 June 2025
- Published by Oxford University Press (OUP) in Systematic Biology
Data Fusion for Integrative Species Identification using Deep Learning
DNA analyses have revolutionized species identification and taxonomic work. Yet, persistent challenges arise from little differentiation among and considerable variation within species, particularly among closely related groups. While images are commonly used as an alternative modality for automated identification tasks, their usability is limited by the same concerns. An integrative strategy, fusing molecular and image data through deep learning, holds significant promise for fine-grained species identification. However, a systematic overview and rigorous statistical testing concerning molecular and image preprocessing and fusion techniques, including practical advice for biologists, are missing so far. We introduce a deep learning scheme that integrates both molecular and image data for species identification. Initially, we systematically assess and compare three different DNA arrangements (aligned, unaligned, single nucleotide polymorphism-reduced) and two encoding methods (fractional, ordinal). Additionally, DL networks are used to extract visual and molecular features, and we propose strategies for fusing this information. Specifically, we investigate three strategies: 1) fusing directly after feature extraction, 2) fusing features that passed through a fully connected layer after feature extraction, and 3) fusing the output scores of both unimodal models. We systematically and statistically evaluate these strategies for 4 eukaryotic datasets, including 2 plant (Asteraceae, Poaceae) and 2 animal families (Lycaenidae, Coccinellidae) using Leave-One-Out Cross-Validation. In addition, we developed an approach to understand molecular- and image-specific identification failure. Aligned sequences with nucleotides encoded using fractional one-hot vectors achieved the highest identification accuracy among DNA data preprocessing techniques in all four datasets. Fusing molecular and visual features directly after feature extraction yielded the best results for three out of four datasets (52–99%). Overall, combining DNA with image data significantly increased accuracy in three out of four datasets, with plant datasets showing the most substantial improvement (Asteraceae: +19%, Poaceae: +13.6%). Even for Lycaenidae with high identification accuracy based on molecular data (>96%), a statistically significant improvement (+2.1%) was observed. Detailed analysis of confusion rates between and within genera shows that DNA alone tends to identify the genus correctly but often fails to recognize the species. The failure to resolve species is alleviated by including image data in the training. This increase in resolution hints at a hierarchical role of modalities in which molecular data coarsely group the specimens to then be guided toward a more fine-grained identification by the connected image. We systematically showed and explained, for the first time, that optimizing the preprocessing and integration of molecular and image data offers significant benefits, particularly for genetically similar and morphologically indistinguishable species, enhancing species identification by reducing modality-specific failure rates and information gaps. Our results can inform integration efforts for various organism groups, improving automated identification across a wide range of eukaryotic species.
- 13 June 2025
- Published by Oxford University Press (OUP) in Systematic Biology
Open AccessData Fusion for Integrative Species Identification using Deep Learning
DNA analyses have revolutionized species identification and taxonomic work. Yet, persistent challenges arise from little differentiation among and considerable variation within species, particularly among closely related groups. While images are commonly used as an alternative modality for automated identification tasks, their usability is limited by the same concerns. An integrative strategy, fusing molecular and image data through deep learning, holds significant promise for fine-grained species identification. However, a systematic overview and rigorous statistical testing concerning molecular and image preprocessing and fusion techniques, including practical advice for biologists, are missing so far. We introduce a deep learning scheme that integrates both molecular and image data for species identification. Initially, we systematically assess and compare three different DNA arrangements (aligned, unaligned, single nucleotide polymorphism-reduced) and two encoding methods (fractional, ordinal). Additionally, DL networks are used to extract visual and molecular features, and we propose strategies for fusing this information. Specifically, we investigate three strategies: 1) fusing directly after feature extraction, 2) fusing features that passed through a fully connected layer after feature extraction, and 3) fusing the output scores of both unimodal models. We systematically and statistically evaluate these strategies for 4 eukaryotic datasets, including 2 plant (Asteraceae, Poaceae) and 2 animal families (Lycaenidae, Coccinellidae) using Leave-One-Out Cross-Validation. In addition, we developed an approach to understand molecular- and image-specific identification failure. Aligned sequences with nucleotides encoded using fractional one-hot vectors achieved the highest identification accuracy among DNA data preprocessing techniques in all four datasets. Fusing molecular and visual features directly after feature extraction yielded the best results for three out of four datasets (52–99%). Overall, combining DNA with image data significantly increased accuracy in three out of four datasets, with plant datasets showing the most substantial improvement (Asteraceae: +19%, Poaceae: +13.6%). Even for Lycaenidae with high identification accuracy based on molecular data (>96%), a statistically significant improvement (+2.1%) was observed. Detailed analysis of confusion rates between and within genera shows that DNA alone tends to identify the genus correctly but often fails to recognize the species. The failure to resolve species is alleviated by including image data in the training. This increase in resolution hints at a hierarchical role of modalities in which molecular data coarsely group the specimens to then be guided toward a more fine-grained identification by the connected image. We systematically showed and explained, for the first time, that optimizing the preprocessing and integration of molecular and image data offers significant benefits, particularly for genetically similar and morphologically indistinguishable species, enhancing species identification by reducing modality-specific failure rates and information gaps. Our results can inform integration efforts for various organism groups, improving automated identification across a wide range of eukaryotic species.
- Open Access
Data Fusion for Integrative Species Identification Using Deep Learning
A bstract: DNA analyses have revolutionized species identification and taxonomic work. Yet, persistent challenges arise from little differentiation among species and considerable variation within species, particularly among closely related groups. While images are commonly used as an alternative modality for automated identification tasks, their usability is limited by the same concerns. An integrative strategy, fusing molecular and image data through machine learning, holds significant promise for fine-grained species identification. However, a systematic overview and rigorous statistical testing concerning molecular and image preprocessing and fusion techniques, including practical advice for biologists, are missing so far. We introduce a machine learning scheme that integrates both molecular and image data for species identification. Initially, we systematically assess and compare three different DNA arrangements (aligned, unaligned, SNP-reduced) and two encoding methods (fractional, ordinal). Later, artificial neural networks are used to extract visual and molecular features, and we propose strategies for fusing this information. Specifically, we investigate three strategies: I) fusing directly after feature extraction, II) fusing features that passed through a fully connected layer after feature extraction, and III) fusing the output scores of both unimodal models. We systematically and statistically evaluate these strategies for four eukaryotic datasets, including two plant (Asteraceae, Poaceae) and two animal families (Lycaenidae, Coccinellidae) using Leave-One-Out Cross-Validation (LOOCV). In addition, we developed an approach to understand molecular-and image-specific identification failure. Aligned sequences with nucleotides encoded as decimal number vectors achieved the highest identification accuracy among DNA data preprocessing techniques in all four datasets. Fusing molecular and visual features directly after feature extraction yielded the best results for three out of four datasets (52-99%). Overall, combining DNA with image data significantly increased accuracy in three out of four datasets, with plant datasets showing the most substantial improvement (Asteraceae: +19%, Poaceae: +13.6%). Even for Lycaenidae with high identification accuracy based on molecular data (>96%), a statistically significant improvement (+2.1%) was observed. Detailed analysis of confusion rates between and within genera shows that DNA alone tends to identify the genus correctly, but often fails to recognize the species. The failure to resolve species is alleviated by including image data in the training. This increase in resolution hints at. a hierarchical role of modalities in which molecular data coarsely groups the specimens to then be guided towards a more fine-grained identification by the connected image. We systematically showed and explained, for the first time, that optimizing the preprocessing and integration of molecular and image data otters significant benefits, particularly for genetically similar and morphologically indistinguishable species, enhancing species identification by reducing modality-specific failure rates and information gaps. Our results can inform integration efforts for various organism groups, improving automated identification across a wide range of eukaryotic species.
Open AccessData Fusion for Integrative Species Identification Using Deep Learning
A bstract: DNA analyses have revolutionized species identification and taxonomic work. Yet, persistent challenges arise from little differentiation among species and considerable variation within species, particularly among closely related groups. While images are commonly used as an alternative modality for automated identification tasks, their usability is limited by the same concerns. An integrative strategy, fusing molecular and image data through machine learning, holds significant promise for fine-grained species identification. However, a systematic overview and rigorous statistical testing concerning molecular and image preprocessing and fusion techniques, including practical advice for biologists, are missing so far. We introduce a machine learning scheme that integrates both molecular and image data for species identification. Initially, we systematically assess and compare three different DNA arrangements (aligned, unaligned, SNP-reduced) and two encoding methods (fractional, ordinal). Later, artificial neural networks are used to extract visual and molecular features, and we propose strategies for fusing this information. Specifically, we investigate three strategies: I) fusing directly after feature extraction, II) fusing features that passed through a fully connected layer after feature extraction, and III) fusing the output scores of both unimodal models. We systematically and statistically evaluate these strategies for four eukaryotic datasets, including two plant (Asteraceae, Poaceae) and two animal families (Lycaenidae, Coccinellidae) using Leave-One-Out Cross-Validation (LOOCV). In addition, we developed an approach to understand molecular-and image-specific identification failure. Aligned sequences with nucleotides encoded as decimal number vectors achieved the highest identification accuracy among DNA data preprocessing techniques in all four datasets. Fusing molecular and visual features directly after feature extraction yielded the best results for three out of four datasets (52-99%). Overall, combining DNA with image data significantly increased accuracy in three out of four datasets, with plant datasets showing the most substantial improvement (Asteraceae: +19%, Poaceae: +13.6%). Even for Lycaenidae with high identification accuracy based on molecular data (>96%), a statistically significant improvement (+2.1%) was observed. Detailed analysis of confusion rates between and within genera shows that DNA alone tends to identify the genus correctly, but often fails to recognize the species. The failure to resolve species is alleviated by including image data in the training. This increase in resolution hints at. a hierarchical role of modalities in which molecular data coarsely groups the specimens to then be guided towards a more fine-grained identification by the connected image. We systematically showed and explained, for the first time, that optimizing the preprocessing and integration of molecular and image data otters significant benefits, particularly for genetically similar and morphologically indistinguishable species, enhancing species identification by reducing modality-specific failure rates and information gaps. Our results can inform integration efforts for various organism groups, improving automated identification across a wide range of eukaryotic species.
- Top Cited PapersOpen Access
- 6 June 2024
- Published by Elsevier in Trends in Ecology & Evolution
- Vol. 39 (8) , 771-784
Species delimitation 4.0: integrative taxonomy meets artificial intelligence
- 6 June 2024
- Published by Elsevier in Trends in Ecology & Evolution
- Vol. 39 (8) , 771-784
Top Cited PapersOpen AccessSpecies delimitation 4.0: integrative taxonomy meets artificial intelligence
- Open Access
- 19 January 2023
- Published by Wiley in Methods in Ecology and Evolution
- Vol. 14 (3) , 799-805
BarKeeper—a versatile web framework to assemble, analyse and manage DNA barcoding data and metadata
DNA barcoding aids in rapidly identifying specimens in various contexts by comparing short stretches of DNA to a reference database. Barcoding initiatives generate large reference databases of carefully curated high‐quality sequences to maximize identification success. BarKeeper provides a flexible tool for barcoding initiatives and in the context of phylogenetic studies to foster shared work on large datasets of raw sequence and associated metadata. It is free, open‐source and available as a set of Docker containers for easy setup. After setting it up once, all project members can use it independently of their operating system or location. BarKeeper offers features to collect and manage data and metadata about specimens, taxa and DNA sequences from Sanger sequencing and high‐throughput sequencing (HTS) technologies. It provides excellent flexibility by not being restricted to specific markers or taxon groups. Users can view and edit records and associated metadata while the app assists them by trimming and assembling reads. It automatically checks the quality of generated barcodes with a taxonomy‐based tool and offers a wide range of options for data analysis. Extensive search features allow querying the database for specific groups of records and saving the search results in the user's profile or downloading them in various file formats. BarKeeper combines multiple tools to aid barcoding projects in every step, from a reference taxa list to finished barcode sequences, thereby minimizing the number of laborious, potentially error‐prone manual steps and enabling efficient collaborative workflows.
- 19 January 2023
- Published by Wiley in Methods in Ecology and Evolution
- Vol. 14 (3) , 799-805
Open AccessBarKeeper—a versatile web framework to assemble, analyse and manage DNA barcoding data and metadata
DNA barcoding aids in rapidly identifying specimens in various contexts by comparing short stretches of DNA to a reference database. Barcoding initiatives generate large reference databases of carefully curated high‐quality sequences to maximize identification success. BarKeeper provides a flexible tool for barcoding initiatives and in the context of phylogenetic studies to foster shared work on large datasets of raw sequence and associated metadata. It is free, open‐source and available as a set of Docker containers for easy setup. After setting it up once, all project members can use it independently of their operating system or location. BarKeeper offers features to collect and manage data and metadata about specimens, taxa and DNA sequences from Sanger sequencing and high‐throughput sequencing (HTS) technologies. It provides excellent flexibility by not being restricted to specific markers or taxon groups. Users can view and edit records and associated metadata while the app assists them by trimming and assembling reads. It automatically checks the quality of generated barcodes with a taxonomy‐based tool and offers a wide range of options for data analysis. Extensive search features allow querying the database for specific groups of records and saving the search results in the user's profile or downloading them in various file formats. BarKeeper combines multiple tools to aid barcoding projects in every step, from a reference taxa list to finished barcode sequences, thereby minimizing the number of laborious, potentially error‐prone manual steps and enabling efficient collaborative workflows.
- Open Access
Gene complementation analysis suggests that dodder plants ( Cuscuta spp.) do not depend on the host FT protein for flowering
Dodder ( Cuscuta spp.) is a genus of parasitic plants that form physiological bridges (haustoria) with their hosts to facilitate the transfer of water and nutrients. The parasites also repurpose nucleic acids and proteins translocating through the haustoria, potentially including the host florigen protein (FT), which is postulated to trigger floral transition in the parasite. Here, we identified the endogenous FT-FD flowering module in Cuscuta campestris . We detected the expression of two parasite-encoded C. campestris (Cc)FT genes in haustoria, whereas a newly found CcFD-like gene was expressed ubiquitously. C. campestris flowered while growing on mutant tobacco plants lacking the floral activators NtFT4 and NtFT5, indicating that host FT proteins are not required to initiate the parasite’s floral transition. We also showed that CcFT1 (identical to CaFT from Cuscuta australis ) and CcFT2 can rescue a non-flowering Ntft4 − Ntft5 − double knockout tobacco phenotype. Together, our results show that Cuscuta spp. produce a potent endogenous florigen as well as other proteins likely to be involved in floral transition. FT gene expression profiles in the haustoria suggest that Cuscuta spp. transition to flowering at least partly in response to host signals (e.g., sugars) that can activate the parasite’s FT-FD module. Although C. campestris and C. australis appear not to depend on the host FT protein for floral transition, the nature of the mobile host signals that influence floral development in these parasites remain unclear. Significance Statement: Parasitic higher plants are known for their sophisticated adaptations that facilitate the transfer of water and nutrients from their hosts. They can also synchronize their transition from vegetative to reproductive development to match the host plant. Despite this high degree of synchronization, dodder plants maintain a potent endogenous floral activator module, which enables the parasite to switch to reproductive development autonomously. Synchronization must therefore involve other stimuli from the host plant, which are currently unknown. Understanding the environmental cues that trigger flowering, and the corresponding network of genetic and physiological regulators and integrators, may lead to new strategies that reduce the reproductive fitness of parasitic plants to protect crops and ensure food security. Data Servers: This article is available as preprint (ID: BIORXIV/2022/520981) at https://www.biorxiv.org under the CC BY-NC 4.0 license. Reusable data files have been deposited at https://datadryad.org , accessible during peer-review under: https://datadryad.org/stash/share/DK8Olh2VqFwbGNL0GtkGt24dD0GhWhJn82oLBC1XK70
Open AccessGene complementation analysis suggests that dodder plants ( Cuscuta spp.) do not depend on the host FT protein for flowering
Dodder ( Cuscuta spp.) is a genus of parasitic plants that form physiological bridges (haustoria) with their hosts to facilitate the transfer of water and nutrients. The parasites also repurpose nucleic acids and proteins translocating through the haustoria, potentially including the host florigen protein (FT), which is postulated to trigger floral transition in the parasite. Here, we identified the endogenous FT-FD flowering module in Cuscuta campestris . We detected the expression of two parasite-encoded C. campestris (Cc)FT genes in haustoria, whereas a newly found CcFD-like gene was expressed ubiquitously. C. campestris flowered while growing on mutant tobacco plants lacking the floral activators NtFT4 and NtFT5, indicating that host FT proteins are not required to initiate the parasite’s floral transition. We also showed that CcFT1 (identical to CaFT from Cuscuta australis ) and CcFT2 can rescue a non-flowering Ntft4 − Ntft5 − double knockout tobacco phenotype. Together, our results show that Cuscuta spp. produce a potent endogenous florigen as well as other proteins likely to be involved in floral transition. FT gene expression profiles in the haustoria suggest that Cuscuta spp. transition to flowering at least partly in response to host signals (e.g., sugars) that can activate the parasite’s FT-FD module. Although C. campestris and C. australis appear not to depend on the host FT protein for floral transition, the nature of the mobile host signals that influence floral development in these parasites remain unclear. Significance Statement: Parasitic higher plants are known for their sophisticated adaptations that facilitate the transfer of water and nutrients from their hosts. They can also synchronize their transition from vegetative to reproductive development to match the host plant. Despite this high degree of synchronization, dodder plants maintain a potent endogenous floral activator module, which enables the parasite to switch to reproductive development autonomously. Synchronization must therefore involve other stimuli from the host plant, which are currently unknown. Understanding the environmental cues that trigger flowering, and the corresponding network of genetic and physiological regulators and integrators, may lead to new strategies that reduce the reproductive fitness of parasitic plants to protect crops and ensure food security. Data Servers: This article is available as preprint (ID: BIORXIV/2022/520981) at https://www.biorxiv.org under the CC BY-NC 4.0 license. Reusable data files have been deposited at https://datadryad.org , accessible during peer-review under: https://datadryad.org/stash/share/DK8Olh2VqFwbGNL0GtkGt24dD0GhWhJn82oLBC1XK70
- 13 January 2021
- Published by Oxford University Press (OUP) in Plant Physiology
- Vol. 185 (4) , 1374-1380
WARPP—web application for the research of parasitic plants
The lifestyle of parasitic plants is associated with peculiar morphological, genetic, and physiological adaptations that existing online plant-specific resources fail to adequately represent. Here, we introduce the Web Application for the Research of Parasitic Plants (WARPP) as an online resource dedicated to advancing research and development of parasitic plant biology. WARPP is a framework to facilitate international efforts by providing a central hub of curated evolutionary, ecological, and genetic data. The first version of WARPP provides a community hub for researchers to test this web application, for which curated data revolving around the economically important Broomrape family (Orobanchaceae) is readily accessible. The initial set of WARPP online tools includes a genome browser that centralizes genomic information for sequenced parasitic plant genomes, an orthogroup summary detailing the presence and absence of orthologous genes in parasites compared with nonparasitic plants, and an ancestral trait explorer showing the evolution of life-history preferences along phylogenies. WARPP represents a project under active development and relies on the scientific community to populate the web app’s database and further the development of new analysis tools. The first version of WARPP can be securely accessed at https://parasiticplants.app. The source code is licensed under GNU GPLv2 and is available at https://github.com/wickeLab/WARPP.
- 13 January 2021
- Published by Oxford University Press (OUP) in Plant Physiology
- Vol. 185 (4) , 1374-1380
WARPP—web application for the research of parasitic plants
The lifestyle of parasitic plants is associated with peculiar morphological, genetic, and physiological adaptations that existing online plant-specific resources fail to adequately represent. Here, we introduce the Web Application for the Research of Parasitic Plants (WARPP) as an online resource dedicated to advancing research and development of parasitic plant biology. WARPP is a framework to facilitate international efforts by providing a central hub of curated evolutionary, ecological, and genetic data. The first version of WARPP provides a community hub for researchers to test this web application, for which curated data revolving around the economically important Broomrape family (Orobanchaceae) is readily accessible. The initial set of WARPP online tools includes a genome browser that centralizes genomic information for sequenced parasitic plant genomes, an orthogroup summary detailing the presence and absence of orthologous genes in parasites compared with nonparasitic plants, and an ancestral trait explorer showing the evolution of life-history preferences along phylogenies. WARPP represents a project under active development and relies on the scientific community to populate the web app’s database and further the development of new analysis tools. The first version of WARPP can be securely accessed at https://parasiticplants.app. The source code is licensed under GNU GPLv2 and is available at https://github.com/wickeLab/WARPP.