Erik Johannes Bekkers
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
Erik Johannes Bekkers has a total of 77 co-authors. Here are the top co-authors by number of shared publications:
- Remco Duits
Department of Mathematics and Computer Science, Eindhoven University of Technology, Eindhoven, The Netherlands
10 shared publications - Alexey Mashtakov
Ailamazyan Program Systems Institute of RAS Pereslavl-Zalessky 152020 Yaroslavl Region Russia
6 shared publications - Bart M. ter Haar Romeny
Eindhoven University of Technology
6 shared publications Remco Duits
Eindhoven University of Technology
6 shared publications- Jiong Zhang
School of Biomedical Engineering, Nanjing University, Suzhou, China
5 shared publications - Behdad Dashtbozorg
Netherlands Cancer Institute
4 shared publications
Latest Publications
- Open Access
- 5 July 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-2
Author Correction: Geometric deep learning improves generalizability of MHC-bound peptide predictions
- 5 July 2025
- Published by Springer Nature in Communications Biology
- Vol. 8 (1) , 1-2
Open AccessAuthor Correction: Geometric deep learning improves generalizability of MHC-bound peptide predictions
- Open Access
Fast and Accurate Peptide–MHC Structure Prediction via an Equivariant Diffusion Model
Accurate modeling of peptide–MHC (Major Histocompatibility Complex) structures is critical for the development of personalized cancer vaccines and T-cell therapies, as MHC proteins present peptides on the cell surface for immune recognition. Here, we introduce MHC-Diff, a specialized SE(3)-equivariant diffusion model leveraging Geometric Deep Learning to predict the 3D C-alpha atom structures of peptide–MHC complexes with high accuracy. Unlike previous deep learning models that generate a single static structure, our probabilistic approach samples multiple diverse candidates, capturing the inherent flexibility of peptide–MHC binding. Validated on the Pandora benchmark and experimental X-ray crystallography data, MHC-Diff achieves sub-angstrom accuracy, outperforming existing methods by a large margin while matching the inference speed of the fastest available techniques. By enabling rapid and highly accurate structure prediction across diverse peptide lengths and MHC alleles, MHC-Diff provides a powerful new tool for accelerating the design of next-generation cancer vaccines and T-cell therapies.
Open AccessFast and Accurate Peptide–MHC Structure Prediction via an Equivariant Diffusion Model
Accurate modeling of peptide–MHC (Major Histocompatibility Complex) structures is critical for the development of personalized cancer vaccines and T-cell therapies, as MHC proteins present peptides on the cell surface for immune recognition. Here, we introduce MHC-Diff, a specialized SE(3)-equivariant diffusion model leveraging Geometric Deep Learning to predict the 3D C-alpha atom structures of peptide–MHC complexes with high accuracy. Unlike previous deep learning models that generate a single static structure, our probabilistic approach samples multiple diverse candidates, capturing the inherent flexibility of peptide–MHC binding. Validated on the Pandora benchmark and experimental X-ray crystallography data, MHC-Diff achieves sub-angstrom accuracy, outperforming existing methods by a large margin while matching the inference speed of the fastest available techniques. By enabling rapid and highly accurate structure prediction across diverse peptide lengths and MHC alleles, MHC-Diff provides a powerful new tool for accelerating the design of next-generation cancer vaccines and T-cell therapies.
- Open Access
- 19 December 2024
- Published by Springer Nature in Communications Biology
- Vol. 7 (1) , 1-12
Geometric deep learning improves generalizability of MHC-bound peptide predictions
The interaction between peptides and major histocompatibility complex (MHC) molecules is pivotal in autoimmunity, pathogen recognition and tumor immunity. Recent advances in cancer immunotherapies demand for more accurate computational prediction of MHC-bound peptides. We address the generalizability challenge of MHC-bound peptide predictions, revealing limitations in current sequence-based approaches. Our structure-based methods leveraging geometric deep learning (GDL) demonstrate promising improvement in generalizability across unseen MHC alleles. Further, we tackle data efficiency by introducing a self-supervised learning approach on structures (3D-SSL). Without being exposed to any binding affinity data, our 3D-SSL outperforms sequence-based methods trained on ~90 times more data points. Finally, we demonstrate the resilience of structure-based GDL methods to biases in binding data on an Hepatitis B virus vaccine immunopeptidomics case study. This proof-of-concept study highlights structure-based methods’ potential to enhance generalizability and data efficiency, with possible implications for data-intensive fields like T-cell receptor specificity predictions.
- 19 December 2024
- Published by Springer Nature in Communications Biology
- Vol. 7 (1) , 1-12
Open AccessGeometric deep learning improves generalizability of MHC-bound peptide predictions
The interaction between peptides and major histocompatibility complex (MHC) molecules is pivotal in autoimmunity, pathogen recognition and tumor immunity. Recent advances in cancer immunotherapies demand for more accurate computational prediction of MHC-bound peptides. We address the generalizability challenge of MHC-bound peptide predictions, revealing limitations in current sequence-based approaches. Our structure-based methods leveraging geometric deep learning (GDL) demonstrate promising improvement in generalizability across unseen MHC alleles. Further, we tackle data efficiency by introducing a self-supervised learning approach on structures (3D-SSL). Without being exposed to any binding affinity data, our 3D-SSL outperforms sequence-based methods trained on ~90 times more data points. Finally, we demonstrate the resilience of structure-based GDL methods to biases in binding data on an Hepatitis B virus vaccine immunopeptidomics case study. This proof-of-concept study highlights structure-based methods’ potential to enhance generalizability and data efficiency, with possible implications for data-intensive fields like T-cell receptor specificity predictions.
- Open Access
Improving generalizability for MHC-I binding peptide predictions through geometric deep learning
Understanding the peptide presentation mechanism of Major Histocompatibility Complex (MHC) is crucial to study the recognition of pathogens, the treatment of autoimmune diseases and in developing cancer immunotherapies. To comprehend these mechanisms and design effective therapies at affordable costs, it is fundamental to be able to precisely predict in silico which peptides can be bound by each MHC. Many computational approaches have been developed to identify MHC-binding peptides based on their sequences, however, they present biases and their performance on less-studied MHC alleles with limited experimental binding data is severely restricted by their sequence-based (SeqB) nature. We reason that structure-based (StrB) methods could learn the physico-chemical and geometrical rules of peptide-MHC (pMHC) complexes conserved among the alleles, thus generalizing better. We designed three supervised StrB geometric deep learning (GDL) methods, showing superior generalization across all StrB methods over two SeqB methods when the test data is distant from the training data. To further explore data efficiency, we designed a novel self-supervised GDL approach trained only on ∼1K x-ray structures, 3D-SSL, which predicts binding affinities without seeing any during training. Finally, we showcase our StrB method on a case study from the HPV vaccine design to show how StrB methods are robust to biases in the binding data. These findings highlight the potentials of structure-based methods in efficiently utilizing limited data, enhancing generalizability. This study bears important implications on data hungry fields, for example, the long-standing T cell receptor (TCR) specificity challenge.
Open AccessImproving generalizability for MHC-I binding peptide predictions through geometric deep learning
Understanding the peptide presentation mechanism of Major Histocompatibility Complex (MHC) is crucial to study the recognition of pathogens, the treatment of autoimmune diseases and in developing cancer immunotherapies. To comprehend these mechanisms and design effective therapies at affordable costs, it is fundamental to be able to precisely predict in silico which peptides can be bound by each MHC. Many computational approaches have been developed to identify MHC-binding peptides based on their sequences, however, they present biases and their performance on less-studied MHC alleles with limited experimental binding data is severely restricted by their sequence-based (SeqB) nature. We reason that structure-based (StrB) methods could learn the physico-chemical and geometrical rules of peptide-MHC (pMHC) complexes conserved among the alleles, thus generalizing better. We designed three supervised StrB geometric deep learning (GDL) methods, showing superior generalization across all StrB methods over two SeqB methods when the test data is distant from the training data. To further explore data efficiency, we designed a novel self-supervised GDL approach trained only on ∼1K x-ray structures, 3D-SSL, which predicts binding affinities without seeing any during training. Finally, we showcase our StrB method on a case study from the HPV vaccine design to show how StrB methods are robust to biases in the binding data. These findings highlight the potentials of structure-based methods in efficiently utilizing limited data, enhancing generalizability. This study bears important implications on data hungry fields, for example, the long-standing T cell receptor (TCR) specificity challenge.
- 1 October 2023
- Published by Springer Nature
- p. 252-261
Regular SE(3) Group Convolutions for Volumetric Medical Image Analysis
- 1 October 2023
- Published by Springer Nature
- p. 252-261
Regular SE(3) Group Convolutions for Volumetric Medical Image Analysis
- Open Access
- 2 August 2023
- Published by Oxford University Press (OUP) in EP Europace
- Vol. 25 (9)
Optimizing patient selection for primary prevention implantable cardioverter-defibrillator implantation: utilizing multimodal machine learning to assess risk of implantable cardioverter-defibrillator non-benefit
Left ventricular ejection fraction (LVEF) is suboptimal as a sole marker for predicting sudden cardiac death (SCD). Machine learning (ML) provides new opportunities for personalized predictions using complex, multimodal data. This study aimed to determine if risk stratification for implantable cardioverter-defibrillator (ICD) implantation can be improved by ML models that combine clinical variables with 12-lead electrocardiograms (ECG) time-series features. A multicentre study of 1010 patients (64.9 ± 10.8 years, 26.8% female) with ischaemic, dilated, or non-ischaemic cardiomyopathy, and LVEF ≤ 35% implanted with an ICD between 2007 and 2021 for primary prevention of SCD in two academic hospitals was performed. For each patient, a raw 12-lead, 10-s ECG was obtained within 90 days before ICD implantation, and clinical details were collected. Supervised ML models were trained and validated on a development cohort (n = 550) from Hospital A to predict ICD non-arrhythmic mortality at three-year follow-up (i.e. mortality without prior appropriate ICD-therapy). Model performance was evaluated on an external patient cohort from Hospital B (n = 460). At three-year follow-up, 16.0% of patients had died, with 72.8% meeting criteria for non-arrhythmic mortality. Extreme gradient boosting models identified patients with non-arrhythmic mortality with an area under the receiver operating characteristic curve (AUROC) of 0.90 [95% confidence intervals (CI) 0.80–1.00] during internal validation. In the external cohort, the AUROC was 0.79 (95% CI 0.75–0.84). ML models combining ECG time-series features and clinical variables were able to predict non-arrhythmic mortality within three years after device implantation in a primary prevention population, with robust performance in an independent cohort.
- 2 August 2023
- Published by Oxford University Press (OUP) in EP Europace
- Vol. 25 (9)
Open AccessOptimizing patient selection for primary prevention implantable cardioverter-defibrillator implantation: utilizing multimodal machine learning to assess risk of implantable cardioverter-defibrillator non-benefit
Left ventricular ejection fraction (LVEF) is suboptimal as a sole marker for predicting sudden cardiac death (SCD). Machine learning (ML) provides new opportunities for personalized predictions using complex, multimodal data. This study aimed to determine if risk stratification for implantable cardioverter-defibrillator (ICD) implantation can be improved by ML models that combine clinical variables with 12-lead electrocardiograms (ECG) time-series features. A multicentre study of 1010 patients (64.9 ± 10.8 years, 26.8% female) with ischaemic, dilated, or non-ischaemic cardiomyopathy, and LVEF ≤ 35% implanted with an ICD between 2007 and 2021 for primary prevention of SCD in two academic hospitals was performed. For each patient, a raw 12-lead, 10-s ECG was obtained within 90 days before ICD implantation, and clinical details were collected. Supervised ML models were trained and validated on a development cohort (n = 550) from Hospital A to predict ICD non-arrhythmic mortality at three-year follow-up (i.e. mortality without prior appropriate ICD-therapy). Model performance was evaluated on an external patient cohort from Hospital B (n = 460). At three-year follow-up, 16.0% of patients had died, with 72.8% meeting criteria for non-arrhythmic mortality. Extreme gradient boosting models identified patients with non-arrhythmic mortality with an area under the receiver operating characteristic curve (AUROC) of 0.90 [95% confidence intervals (CI) 0.80–1.00] during internal validation. In the external cohort, the AUROC was 0.79 (95% CI 0.75–0.84). ML models combining ECG time-series features and clinical variables were able to predict non-arrhythmic mortality within three years after device implantation in a primary prevention population, with robust performance in an independent cohort.
- Open Access
B-Spline CNNs on Lie Groups
Group convolutional neural networks (G-CNNs) can be used to improve classical CNNs by equipping them with the geometric structure of groups. Central in the success of G-CNNs is the lifting of feature maps to higher dimensional disentangled representations, in which data characteristics are effectively learned, geometric data-augmentations are made obsolete, and predictable behavior under geometric transformations (equivariance) is guaranteed via group theory. Currently, however, the practical implementations of G-CNNs are limited to either discrete groups (that leave the grid intact) or continuous compact groups such as rotations (that enable the use of Fourier theory). In this paper we lift these limitations and propose a modular framework for the design and implementation of G-CNNs for arbitrary Lie groups. In our approach the differential structure of Lie groups is used to expand convolution kernels in a generic basis of B-splines that is defined on the Lie algebra. This leads to a flexible framework that enables localized, atrous, and deformable convolutions in G-CNNs by means of respectively localized, sparse and non-uniform B-spline expansions. The impact and potential of our approach is studied on two benchmark datasets: cancer detection in histopathology slides in which rotation equivariance plays a key role and facial landmark localization in which scale equivariance is important. In both cases, G-CNN architectures outperform their classical 2D counterparts and the added value of atrous and localized group convolutions is studied in detail.
Open AccessB-Spline CNNs on Lie Groups
Group convolutional neural networks (G-CNNs) can be used to improve classical CNNs by equipping them with the geometric structure of groups. Central in the success of G-CNNs is the lifting of feature maps to higher dimensional disentangled representations, in which data characteristics are effectively learned, geometric data-augmentations are made obsolete, and predictable behavior under geometric transformations (equivariance) is guaranteed via group theory. Currently, however, the practical implementations of G-CNNs are limited to either discrete groups (that leave the grid intact) or continuous compact groups such as rotations (that enable the use of Fourier theory). In this paper we lift these limitations and propose a modular framework for the design and implementation of G-CNNs for arbitrary Lie groups. In our approach the differential structure of Lie groups is used to expand convolution kernels in a generic basis of B-splines that is defined on the Lie algebra. This leads to a flexible framework that enables localized, atrous, and deformable convolutions in G-CNNs by means of respectively localized, sparse and non-uniform B-spline expansions. The impact and potential of our approach is studied on two benchmark datasets: cancer detection in histopathology slides in which rotation equivariance plays a key role and facial landmark localization in which scale equivariance is important. In both cases, G-CNN architectures outperform their classical 2D counterparts and the added value of atrous and localized group convolutions is studied in detail.
- Open Access
Fourier Transform on the Homogeneous Space of 3D Positions and Orientations for Exact Solutions to Linear PDEs
Fokker–Planck PDEs (including diffusions) for stable Lévy processes (including Wiener processes) on the joint space of positions and orientations play a major role in mechanics, robotics, image analysis, directional statistics and probability theory. Exact analytic designs and solutions are known in the 2D case, where they have been obtained using Fourier transform on . Here, we extend these approaches to 3D using Fourier transform on the Lie group of rigid body motions. More precisely, we define the homogeneous space of 3D positions and orientations as the quotient in . In our construction, two group elements are equivalent if they are equal up to a rotation around the reference axis. On this quotient, we design a specific Fourier transform. We apply this Fourier transform to derive new exact solutions to Fokker–Planck PDEs of -stable Lévy processes on . This reduces classical analysis computations and provides an explicit algebraic spectral decomposition of the solutions. We compare the exact probability kernel for (the diffusion kernel) to the kernel for (the Poisson kernel). We set up stochastic differential equations (SDEs) for the Lévy processes on the quotient and derive corresponding Monte-Carlo methods. We verified that the exact probability kernels arise as the limit of the Monte-Carlo approximations.Open AccessFourier Transform on the Homogeneous Space of 3D Positions and Orientations for Exact Solutions to Linear PDEs
Fokker–Planck PDEs (including diffusions) for stable Lévy processes (including Wiener processes) on the joint space of positions and orientations play a major role in mechanics, robotics, image analysis, directional statistics and probability theory. Exact analytic designs and solutions are known in the 2D case, where they have been obtained using Fourier transform on . Here, we extend these approaches to 3D using Fourier transform on the Lie group of rigid body motions. More precisely, we define the homogeneous space of 3D positions and orientations as the quotient in . In our construction, two group elements are equivalent if they are equal up to a rotation around the reference axis. On this quotient, we design a specific Fourier transform. We apply this Fourier transform to derive new exact solutions to Fokker–Planck PDEs of -stable Lévy processes on . This reduces classical analysis computations and provides an explicit algebraic spectral decomposition of the solutions. We compare the exact probability kernel for (the diffusion kernel) to the kernel for (the Poisson kernel). We set up stochastic differential equations (SDEs) for the Lévy processes on the quotient and derive corresponding Monte-Carlo methods. We verified that the exact probability kernels arise as the limit of the Monte-Carlo approximations. - 26 September 2018
- Published by Springer Nature
- p. 440-448
Roto-Translation Covariant Convolutional Networks for Medical Image Analysis
- 26 September 2018
- Published by Springer Nature
- p. 440-448
Roto-Translation Covariant Convolutional Networks for Medical Image Analysis
- Open Access
Roto-Translation Covariant Convolutional Networks for Medical Image Analysis
We propose a framework for rotation and translation covariant deep learning using $SE(2)$ group convolutions. The group product of the special Euclidean motion group $SE(2)$ describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via $SE(2)$ group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an $SE(2)$-image, i.e., 3D (vector valued) data whose domain is $SE(2)$; a group convolution layer from and to an $SE(2)$-image; and a projection layer from an $SE(2)$-image to a 2D image. The lifting and group convolution layers are $SE(2)$ covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.Open AccessRoto-Translation Covariant Convolutional Networks for Medical Image Analysis
We propose a framework for rotation and translation covariant deep learning using $SE(2)$ group convolutions. The group product of the special Euclidean motion group $SE(2)$ describes how a concatenation of two roto-translations results in a net roto-translation. We encode this geometric structure into convolutional neural networks (CNNs) via $SE(2)$ group convolutional layers, which fit into the standard 2D CNN framework, and which allow to generically deal with rotated input samples without the need for data augmentation. We introduce three layers: a lifting layer which lifts a 2D (vector valued) image to an $SE(2)$-image, i.e., 3D (vector valued) data whose domain is $SE(2)$; a group convolution layer from and to an $SE(2)$-image; and a projection layer from an $SE(2)$-image to a 2D image. The lifting and group convolution layers are $SE(2)$ covariant (the output roto-translates with the input). The final projection layer, a maximum intensity projection over rotations, makes the full CNN rotation invariant. We show with three different problems in histopathology, retinal imaging, and electron microscopy that with the proposed group CNNs, state-of-the-art performance can be achieved, without the need for data augmentation by rotation and with increased performance compared to standard CNNs that do rely on augmentation.