Philipp Marquardt
Research Subjects
Philipp Marquardt has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Alexander Nesterov-Mueller
Institute of Microstructure Technology, Karlsruhe Institute for Technology, Hermann-von-Helmholtz-Platz 1, 76344 Eggenstein-Leopoldshafen, Germany
1 shared publication - Anja Seckinger
Department of Hematology and Immunology, Myeloma Center Brussels & Labor für Myelomforschung, Vrije Universiteit Brussel (VUB), Jette, Belgium
1 shared publication - Anne Rübsam
Charité – Universitätsmedizin Berlin, Freie Universität Berlin and Humboldt- Universität zu Berlin
1 shared publication - Antonia Maria Joussen
Department of Ophthalmology, Charité—Universitätsmedizin Berlin, 12203 Berlin, Germany
1 shared publication - Clemens von Bojničić-Kninski
Axxelera, Karl-Floesser-Str. 14, 76189 Karlsruhe, Germany
1 shared publication - Constance B. Runkel
Department of Ophthalmology, Charité University Hospital Berlin, 12203 Berlin, Germany
1 shared publication
Latest Publications
- Open Access
- 16 July 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (14) , 4141
Using Deep Learning to Distinguish Highly Malignant Uveal Melanoma from Benign Choroidal Nevi
Background: This study aimed to evaluate the potential of human–machine interaction (HMI) in a deep learning software for discerning the malignancy of choroidal melanocytic lesions based on fundus photographs. Methods: The study enrolled individuals diagnosed with a choroidal melanocytic lesion at a tertiary clinic between 2011 and 2023, resulting in a cohort of 762 eligible cases. A deep learning-based assistant integrated into the software underwent training using a dataset comprising 762 color fundus photographs (CFPs) of choroidal lesions captured by various fundus cameras. The dataset was categorized into benign nevi, untreated choroidal melanomas, and irradiated choroidal melanomas. The reference standard for evaluation was established by retinal specialists using multimodal imaging. Trinary and binary models were trained, and their classification performance was evaluated on a test set consisting of 100 independent images. The discriminative performance of deep learning models was evaluated based on accuracy, recall, and specificity. Results: The final accuracy rates on the independent test set for multi-class and binary (benign vs. malignant) classification were 84.8% and 90.9%, respectively. Recall and specificity ranged from 0.85 to 0.90 and 0.91 to 0.92, respectively. The mean area under the curve (AUC) values were 0.96 and 0.99, respectively. Optimal discriminative performance was observed in binary classification with the incorporation of a single imaging modality, achieving an accuracy of 95.8%. Conclusions: The deep learning models demonstrated commendable performance in distinguishing the malignancy of choroidal lesions. The software exhibits promise for resource-efficient and cost-effective pre-stratification.
- 16 July 2024
- Published by MDPI AG in Journal of Clinical Medicine
- Vol. 13 (14) , 4141
Open AccessUsing Deep Learning to Distinguish Highly Malignant Uveal Melanoma from Benign Choroidal Nevi
Background: This study aimed to evaluate the potential of human–machine interaction (HMI) in a deep learning software for discerning the malignancy of choroidal melanocytic lesions based on fundus photographs. Methods: The study enrolled individuals diagnosed with a choroidal melanocytic lesion at a tertiary clinic between 2011 and 2023, resulting in a cohort of 762 eligible cases. A deep learning-based assistant integrated into the software underwent training using a dataset comprising 762 color fundus photographs (CFPs) of choroidal lesions captured by various fundus cameras. The dataset was categorized into benign nevi, untreated choroidal melanomas, and irradiated choroidal melanomas. The reference standard for evaluation was established by retinal specialists using multimodal imaging. Trinary and binary models were trained, and their classification performance was evaluated on a test set consisting of 100 independent images. The discriminative performance of deep learning models was evaluated based on accuracy, recall, and specificity. Results: The final accuracy rates on the independent test set for multi-class and binary (benign vs. malignant) classification were 84.8% and 90.9%, respectively. Recall and specificity ranged from 0.85 to 0.90 and 0.91 to 0.92, respectively. The mean area under the curve (AUC) values were 0.96 and 0.99, respectively. Optimal discriminative performance was observed in binary classification with the incorporation of a single imaging modality, achieving an accuracy of 95.8%. Conclusions: The deep learning models demonstrated commendable performance in distinguishing the malignancy of choroidal lesions. The software exhibits promise for resource-efficient and cost-effective pre-stratification.
- Open Access
- 23 March 2022
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 23 (7) , 3515
Resemblance-Ranking Peptide Library to Screen for Binders to Antibodies on a Peptidomic Scale
A novel resemblance-ranking peptide library with 160,000 10-meric peptides was designed to search for selective binders to antibodies. The resemblance-ranking principle enabled the selection of sequences that are most similar to the human peptidome. The library was synthesized with ultra-high-density peptide arrays. As proof of principle, screens for selective binders were performed for the therapeutic anti-CD20 antibody rituximab. Several features in the amino acid composition of antibody-binding peptides were identified. The selective affinity of rituximab increased with an increase in the number of hydrophobic amino acids in a peptide, mainly tryptophan and phenylalanine, while a total charge of the peptide remained relatively small. Peptides with a higher affinity exhibited a lower sum helix propensity. For the 30 strongest peptide binders, a substitutional analysis was performed to determine dissociation constants and the invariant amino acids for binding to rituximab. The strongest selective peptides had a dissociation constant in the hundreds of the nano-molar range. The substitutional analysis revealed a specific hydrophobic epitope for rituximab. To show that conformational binders can, in principle, be detected in array format, cyclic peptide substitutions that are similar to the target of rituximab were investigated. Since the specific binders selected via the resemblance-ranking peptide library were based on the hydrophobic interactions that are widespread in the world of biomolecules, the library can be used to screen for potential linear epitopes that may provide information about the cross-reactivity of antibodies.
- 23 March 2022
- Published by MDPI AG in International Journal of Molecular Sciences
- Vol. 23 (7) , 3515
Open AccessResemblance-Ranking Peptide Library to Screen for Binders to Antibodies on a Peptidomic Scale
A novel resemblance-ranking peptide library with 160,000 10-meric peptides was designed to search for selective binders to antibodies. The resemblance-ranking principle enabled the selection of sequences that are most similar to the human peptidome. The library was synthesized with ultra-high-density peptide arrays. As proof of principle, screens for selective binders were performed for the therapeutic anti-CD20 antibody rituximab. Several features in the amino acid composition of antibody-binding peptides were identified. The selective affinity of rituximab increased with an increase in the number of hydrophobic amino acids in a peptide, mainly tryptophan and phenylalanine, while a total charge of the peptide remained relatively small. Peptides with a higher affinity exhibited a lower sum helix propensity. For the 30 strongest peptide binders, a substitutional analysis was performed to determine dissociation constants and the invariant amino acids for binding to rituximab. The strongest selective peptides had a dissociation constant in the hundreds of the nano-molar range. The substitutional analysis revealed a specific hydrophobic epitope for rituximab. To show that conformational binders can, in principle, be detected in array format, cyclic peptide substitutions that are similar to the target of rituximab were investigated. Since the specific binders selected via the resemblance-ranking peptide library were based on the hydrophobic interactions that are widespread in the world of biomolecules, the library can be used to screen for potential linear epitopes that may provide information about the cross-reactivity of antibodies.