Karim Beguir
Research Subjects
Karim Beguir has a total of 88 co-authors. Here are the top co-authors by number of shared publications:
- Yunguan Fu
InstaDeep Ltd
6 shared publications - Nicolas Lopez Carranza
InstaDeep Ltd, London, UK
5 shared publications - Marcin J. Skwark
VetAI.bio Ltd., London UB9 5NJ, UK
4 shared publications - Ugur Sahin
BioNTech SE, Mainz, Germany
4 shared publications - Maren Lang
BioNTech, Mainz, Germany
3 shared publications - Thomas Pierrot
InstaDeep
3 shared publications
Latest Publications
- Open Access
FrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting
Protein structure prediction field has been revolutionised by deep learning with protein folding models such as AlphaFold 2 and ESMFold. These models enable rapid in silico prediction and have been integrated into de novo protein design and protein-protein interaction (PPI) prediction. However, biologically relevant features dependent on conformational distributions cannot be estimated with these models. Diffusion models, a novel class of generative models, have been developed to learn conformational distributions and applied to de novo protein design. Limited work has been done on protein structure inpainting, where a masked section is recovered by simultaneously conditioning on its sequence and the rest of the structure. In this work, we propose FrameD iff i n P ain T ing (FrameDiPT), a generalised model for protein inpainting. This is important for T-cells given the hyper-variability of the complementarity determining region (CDR) loops. We evaluated the model on CDR loop design for T-cell receptors and achieved comparable prediction accuracy to ProteinGenerator and RFdiffusion with limited training data and learnable parameters. Different from deterministic structure prediction models, FrameDiPT captures the conformational distribution at different regions and binding states, highlighting a key advantage of generative models. The model and inference code have been released 1 .
Open AccessFrameDiPT: SE(3) Diffusion Model for Protein Structure Inpainting
Protein structure prediction field has been revolutionised by deep learning with protein folding models such as AlphaFold 2 and ESMFold. These models enable rapid in silico prediction and have been integrated into de novo protein design and protein-protein interaction (PPI) prediction. However, biologically relevant features dependent on conformational distributions cannot be estimated with these models. Diffusion models, a novel class of generative models, have been developed to learn conformational distributions and applied to de novo protein design. Limited work has been done on protein structure inpainting, where a masked section is recovered by simultaneously conditioning on its sequence and the rest of the structure. In this work, we propose FrameD iff i n P ain T ing (FrameDiPT), a generalised model for protein inpainting. This is important for T-cells given the hyper-variability of the complementarity determining region (CDR) loops. We evaluated the model on CDR loop design for T-cell receptors and achieved comparable prediction accuracy to ProteinGenerator and RFdiffusion with limited training data and learnable parameters. Different from deterministic structure prediction models, FrameDiPT captures the conformational distribution at different regions and binding states, highlighting a key advantage of generative models. The model and inference code have been released 1 .
- Open Access
LightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks
The peptide-major histocompatibility complex (pMHC) is a crucial protein in cell-mediated immune recognition and response. Accurate structure prediction is potentially beneficial for protein interaction prediction and therefore helps immunotherapy design. However, predicting these structures is challenging due to the sequential and structural variability. In addition, existing pre-trained models such as AlphaFold 2 require expensive computation thus inhibiting high throughput in silico peptide screening. In this study, we propose LightMHC: a lightweight model (2.2M parameters) equipped with attention mechanisms, graph neural networks, and convolutional neural networks. LightMHC predicts full-atom pMHC structures from amino-acid sequences alone, without template structures. The model achieved comparable or superior performance to AlphaFold 2 and ESMFold (93M and 15B parameters respectively), with five-fold acceleration (6.65 seconds/sample for LightMHC versus 36.82 seconds/sample for AlphaFold 2), potentially offering a valuable tool for immune protein structure prediction and immunotherapy design.
Open AccessLightMHC: A Light Model for pMHC Structure Prediction with Graph Neural Networks
The peptide-major histocompatibility complex (pMHC) is a crucial protein in cell-mediated immune recognition and response. Accurate structure prediction is potentially beneficial for protein interaction prediction and therefore helps immunotherapy design. However, predicting these structures is challenging due to the sequential and structural variability. In addition, existing pre-trained models such as AlphaFold 2 require expensive computation thus inhibiting high throughput in silico peptide screening. In this study, we propose LightMHC: a lightweight model (2.2M parameters) equipped with attention mechanisms, graph neural networks, and convolutional neural networks. LightMHC predicts full-atom pMHC structures from amino-acid sequences alone, without template structures. The model achieved comparable or superior performance to AlphaFold 2 and ESMFold (93M and 15B parameters respectively), with five-fold acceleration (6.65 seconds/sample for LightMHC versus 36.82 seconds/sample for AlphaFold 2), potentially offering a valuable tool for immune protein structure prediction and immunotherapy design.
- Open Access
A Foundational Large Language Model for Edible Plant Genomes
Significant progress has been made in the field of plant genomics, as demonstrated by the increased use of high-throughput methodologies that enable the characterization of multiple genome-wide molecular phenotypes. These findings have provided valuable insights into plant traits and their underlying genetic mechanisms, particularly in model plant species. Nonetheless, effectively leveraging them to make accurate predictions represents a critical step in crop genomic improvement. We present AgroNT, a foundational large language model trained on genomes from 48 plant species with a predominant focus on crop species. We show that AgroNT can obtain state-of-the-art predictions for regulatory annotations, promoter/terminator strength, tissue-specific gene expression, and prioritize functional variants. We conduct a large-scale in silico saturation mutagenesis analysis on cassava to evaluate the regulatory impact of over 10 million mutations and provide their predicted effects as a resource for variant characterization. Finally, we propose the use of the diverse datasets compiled here as the Plants Genomic Benchmark (PGB), providing a comprehensive benchmark for deep learning-based methods in plant genomic research. The pre-trained AgroNT model is publicly available on HuggingFace at https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b for future research purposes.
Open AccessA Foundational Large Language Model for Edible Plant Genomes
Significant progress has been made in the field of plant genomics, as demonstrated by the increased use of high-throughput methodologies that enable the characterization of multiple genome-wide molecular phenotypes. These findings have provided valuable insights into plant traits and their underlying genetic mechanisms, particularly in model plant species. Nonetheless, effectively leveraging them to make accurate predictions represents a critical step in crop genomic improvement. We present AgroNT, a foundational large language model trained on genomes from 48 plant species with a predominant focus on crop species. We show that AgroNT can obtain state-of-the-art predictions for regulatory annotations, promoter/terminator strength, tissue-specific gene expression, and prioritize functional variants. We conduct a large-scale in silico saturation mutagenesis analysis on cassava to evaluate the regulatory impact of over 10 million mutations and provide their predicted effects as a resource for variant characterization. Finally, we propose the use of the diverse datasets compiled here as the Plants Genomic Benchmark (PGB), providing a comprehensive benchmark for deep learning-based methods in plant genomic research. The pre-trained AgroNT model is publicly available on HuggingFace at https://huggingface.co/InstaDeepAI/agro-nucleotide-transformer-1b for future research purposes.
- Open Access
De novo peptide sequencing with InstaNovo: Accurate, database-free peptide identification for large scale proteomics experiments
Bottom-up mass spectrometry-based proteomics is challenged by the task of identifying the peptide that generates a tandem mass spectrum. Traditional methods that rely on known peptide sequence databases are limited and may not be applicable in certain contexts. De novo peptide sequencing, which assigns peptide sequences to the spectra without prior information, is valuable for various biological applications; yet, due to a lack of accuracy, it remains challenging to apply this approach in many situations. Here, we introduce InstaNovo, a transformer neural network with the ability to translate fragment ion peaks into the sequence of amino acids that make up the studied peptide(s). The model was trained on 28 million labelled spectra matched to 742k human peptides from the ProteomeTools project. We demonstrate that InstaNovo outperforms current state-of-the-art methods on benchmark datasets and showcase its utility in several applications. Building upon human intuition, we also introduce InstaNovo+, a multinomial diffusion model that further improves performance by iterative refinement of predicted sequences. Using these models, we could de novo sequence antibody-based therapeutics with unprecedented coverage, discover novel peptides, and detect unreported organisms in different datasets, thereby expanding the scope and detection rate of proteomics searches. Finally, we could experimentally validate tryptic and non-tryptic peptides with targeted proteomics, demonstrating the fidelity of our predictions. Our models unlock a plethora of opportunities across different scientific domains, such as direct protein sequencing, immunopeptidomics, and exploration of the dark proteome.
Open AccessDe novo peptide sequencing with InstaNovo: Accurate, database-free peptide identification for large scale proteomics experiments
Bottom-up mass spectrometry-based proteomics is challenged by the task of identifying the peptide that generates a tandem mass spectrum. Traditional methods that rely on known peptide sequence databases are limited and may not be applicable in certain contexts. De novo peptide sequencing, which assigns peptide sequences to the spectra without prior information, is valuable for various biological applications; yet, due to a lack of accuracy, it remains challenging to apply this approach in many situations. Here, we introduce InstaNovo, a transformer neural network with the ability to translate fragment ion peaks into the sequence of amino acids that make up the studied peptide(s). The model was trained on 28 million labelled spectra matched to 742k human peptides from the ProteomeTools project. We demonstrate that InstaNovo outperforms current state-of-the-art methods on benchmark datasets and showcase its utility in several applications. Building upon human intuition, we also introduce InstaNovo+, a multinomial diffusion model that further improves performance by iterative refinement of predicted sequences. Using these models, we could de novo sequence antibody-based therapeutics with unprecedented coverage, discover novel peptides, and detect unreported organisms in different datasets, thereby expanding the scope and detection rate of proteomics searches. Finally, we could experimentally validate tryptic and non-tryptic peptides with targeted proteomics, demonstrating the fidelity of our predictions. Our models unlock a plethora of opportunities across different scientific domains, such as direct protein sequencing, immunopeptidomics, and exploration of the dark proteome.
- Open Access
- 2 February 2023
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 155, 106618
Early computational detection of potential high-risk SARS-CoV-2 variants
- 2 February 2023
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 155, 106618
Open AccessEarly computational detection of potential high-risk SARS-CoV-2 variants
- Open Access
Peptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network
Neoantigen-targeting vaccines have achieved breakthrough success in cancer immunotherapy by eliciting immune responses against neoantigens, which are proteins uniquely produced by cancer cells. During the immune response, the interactions between peptides and major histocompatibility complexes (MHC) play an important role as peptides must be bound and presented by MHC to be recognised by the immune system. However, only limited experimentally determined peptide-MHC (pMHC) structures are available, and in-silico structure modelling is therefore used for studying their interactions. Current approaches mainly use Monte Carlo sampling and energy minimisation, and are often computationally expensive. On the other hand, the advent of large high-quality proteomic data sets has led to an unprecedented opportunity for deep learning-based methods with pMHC structure prediction becoming feasible with these trained protein folding models. In this work, we present a graph neural network-based model for pMHC structure prediction, which takes an amino acid-level pMHC graph and an atomic-level peptide graph as inputs and predicts the peptide backbone conformation. With a novel weighted reconstruction loss, the trained model achieved a similar accuracy to AlphaFold 2, requiring only 1.7M learnable parameters compared to 93M, representing a more than 98% reduction in the number of required parameters.
Open AccessPeptide-MHC Structure Prediction With Mixed Residue and Atom Graph Neural Network
Neoantigen-targeting vaccines have achieved breakthrough success in cancer immunotherapy by eliciting immune responses against neoantigens, which are proteins uniquely produced by cancer cells. During the immune response, the interactions between peptides and major histocompatibility complexes (MHC) play an important role as peptides must be bound and presented by MHC to be recognised by the immune system. However, only limited experimentally determined peptide-MHC (pMHC) structures are available, and in-silico structure modelling is therefore used for studying their interactions. Current approaches mainly use Monte Carlo sampling and energy minimisation, and are often computationally expensive. On the other hand, the advent of large high-quality proteomic data sets has led to an unprecedented opportunity for deep learning-based methods with pMHC structure prediction becoming feasible with these trained protein folding models. In this work, we present a graph neural network-based model for pMHC structure prediction, which takes an amino acid-level pMHC graph and an atomic-level peptide graph as inputs and predicts the peptide backbone conformation. With a novel weighted reconstruction loss, the trained model achieved a similar accuracy to AlphaFold 2, requiring only 1.7M learnable parameters compared to 93M, representing a more than 98% reduction in the number of required parameters.
- Open Access
So ManyFolds, So Little Time: Efficient Protein Structure Prediction With pLMs and MSAs
In recent years, machine learning approaches for de novo protein structure prediction have made significant progress, culminating in AlphaFold which approaches experimental accuracies in certain settings and heralds the possibility of rapid in silico protein modelling and design. However, such applications can be challenging in practice due to the significant compute required for training and inference of such models, and their strong reliance on the evolutionary information contained in multiple sequence alignments (MSAs), which may not be available for certain targets of interest. Here, we first present a streamlined AlphaFold architecture and training pipeline that still provides good performance with significantly reduced computational burden. Aligned with recent approaches such as OmegaFold and ESMFold, our model is initially trained to predict structure from sequences alone by leveraging embeddings from the pretrained ESM-2 protein language model (pLM). We then compare this approach to an equivalent model trained on MSA-profile information only, and find that the latter still provides a performance boost – suggesting that even state-of-the-art pLMs cannot yet easily replace the evolutionary information of homologous sequences. Finally, we train a model that can make predictions from either the combination, or only one, of pLM and MSA inputs. Ultimately, we obtain accuracies in any of these three input modes similar to models trained uniquely in that setting, whilst also demonstrating that these modalities are complimentary, each regularly outperforming the other.
Open AccessSo ManyFolds, So Little Time: Efficient Protein Structure Prediction With pLMs and MSAs
In recent years, machine learning approaches for de novo protein structure prediction have made significant progress, culminating in AlphaFold which approaches experimental accuracies in certain settings and heralds the possibility of rapid in silico protein modelling and design. However, such applications can be challenging in practice due to the significant compute required for training and inference of such models, and their strong reliance on the evolutionary information contained in multiple sequence alignments (MSAs), which may not be available for certain targets of interest. Here, we first present a streamlined AlphaFold architecture and training pipeline that still provides good performance with significantly reduced computational burden. Aligned with recent approaches such as OmegaFold and ESMFold, our model is initially trained to predict structure from sequences alone by leveraging embeddings from the pretrained ESM-2 protein language model (pLM). We then compare this approach to an equivalent model trained on MSA-profile information only, and find that the latter still provides a performance boost – suggesting that even state-of-the-art pLMs cannot yet easily replace the evolutionary information of homologous sequences. Finally, we train a model that can make predictions from either the combination, or only one, of pLM and MSA inputs. Ultimately, we obtain accuracies in any of these three input modes similar to models trained uniquely in that setting, whilst also demonstrating that these modalities are complimentary, each regularly outperforming the other.
- Open Access
Early Computational Detection of Potential High Risk SARS-CoV-2 Variants
The ongoing COVID-19 pandemic is leading to the discovery of hundreds of novel SARS-CoV-2 variants on a daily basis. While most variants do not impact the course of the pandemic, some variants pose a significantly increased risk when the acquired mutations allow better evasion of antibody neutralisation in previously infected or vaccinated subjects or increased transmissibility. Early detection of such high risk variants (HRVs) is paramount for the proper management of the pandemic. However, experimental assays to determine immune evasion and transmissibility characteristics of new variants are resource-intensive and time-consuming, potentially leading to delays in appropriate responses by decision makers. Here we present a novel in silico approach combining spike (S) protein structure modelling and large protein transformer language models on S protein sequences to accurately rank SARS-CoV-2 variants for immune escape and fitness potential. These metrics can be combined into an automated Early Warning System (EWS) capable of evaluating new variants in minutes and risk-monitoring variant lineages in near real-time. The system accurately pinpoints the putatively dangerous variants by selecting on average less than 0.3% of the novel variants each week. With only the S protein nucleotide sequence as input, the EWS detects HRVs earlier and with better precision than baseline metrics such as the growth metric (which requires real-world observations) or random sampling. Notably, Omicron BA.1 was flagged by the EWS on the day its sequence was made available. Additionally, our immune escape and fitness metrics were experimentally validated using in vitro pseudovirus-based virus neutralisation test (pVNT) assays and binding assays. The EWS flagged as potentially dangerous all 16 variants (Alpha-Omicron BA.1/2/4/5) designated by the World Health Organisation (WHO) with an average lead time of more than one and a half months ahead of them being designated as such. One-Sentence Summary: A COVID-19 Early Warning System combining structural modelling with machine learning to detect and monitor high risk SARS-CoV-2 variants, identifying all 16 WHO designated variants on average more than one and a half months in advance by selecting on average less than 0.3% of the weekly novel variants.
Open AccessEarly Computational Detection of Potential High Risk SARS-CoV-2 Variants
The ongoing COVID-19 pandemic is leading to the discovery of hundreds of novel SARS-CoV-2 variants on a daily basis. While most variants do not impact the course of the pandemic, some variants pose a significantly increased risk when the acquired mutations allow better evasion of antibody neutralisation in previously infected or vaccinated subjects or increased transmissibility. Early detection of such high risk variants (HRVs) is paramount for the proper management of the pandemic. However, experimental assays to determine immune evasion and transmissibility characteristics of new variants are resource-intensive and time-consuming, potentially leading to delays in appropriate responses by decision makers. Here we present a novel in silico approach combining spike (S) protein structure modelling and large protein transformer language models on S protein sequences to accurately rank SARS-CoV-2 variants for immune escape and fitness potential. These metrics can be combined into an automated Early Warning System (EWS) capable of evaluating new variants in minutes and risk-monitoring variant lineages in near real-time. The system accurately pinpoints the putatively dangerous variants by selecting on average less than 0.3% of the novel variants each week. With only the S protein nucleotide sequence as input, the EWS detects HRVs earlier and with better precision than baseline metrics such as the growth metric (which requires real-world observations) or random sampling. Notably, Omicron BA.1 was flagged by the EWS on the day its sequence was made available. Additionally, our immune escape and fitness metrics were experimentally validated using in vitro pseudovirus-based virus neutralisation test (pVNT) assays and binding assays. The EWS flagged as potentially dangerous all 16 variants (Alpha-Omicron BA.1/2/4/5) designated by the World Health Organisation (WHO) with an average lead time of more than one and a half months ahead of them being designated as such. One-Sentence Summary: A COVID-19 Early Warning System combining structural modelling with machine learning to detect and monitor high risk SARS-CoV-2 variants, identifying all 16 WHO designated variants on average more than one and a half months in advance by selecting on average less than 0.3% of the weekly novel variants.
- Open Access
Learning Compositional Neural Programs for Continuous Control
We propose a novel solution to challenging sparse-reward, continuous control problems that require hierarchical planning at multiple levels of abstraction. Our solution, dubbed AlphaNPI-X, involves three separate stages of learning. First, we use off-policy reinforcement learning algorithms with experience replay to learn a set of atomic goal-conditioned policies, which can be easily repurposed for many tasks. Second, we learn self-models describing the effect of the atomic policies on the environment. Third, the self-models are harnessed to learn recursive compositional programs with multiple levels of abstraction. The key insight is that the self-models enable planning by imagination, obviating the need for interaction with the world when learning higher-level compositional programs. To accomplish the third stage of learning, we extend the AlphaNPI algorithm, which applies AlphaZero to learn recursive neural programmer-interpreters. We empirically show that AlphaNPI-X can effectively learn to tackle challenging sparse manipulation tasks, such as stacking multiple blocks, where powerful model-free baselines fail.
Open AccessLearning Compositional Neural Programs for Continuous Control
We propose a novel solution to challenging sparse-reward, continuous control problems that require hierarchical planning at multiple levels of abstraction. Our solution, dubbed AlphaNPI-X, involves three separate stages of learning. First, we use off-policy reinforcement learning algorithms with experience replay to learn a set of atomic goal-conditioned policies, which can be easily repurposed for many tasks. Second, we learn self-models describing the effect of the atomic policies on the environment. Third, the self-models are harnessed to learn recursive compositional programs with multiple levels of abstraction. The key insight is that the self-models enable planning by imagination, obviating the need for interaction with the world when learning higher-level compositional programs. To accomplish the third stage of learning, we extend the AlphaNPI algorithm, which applies AlphaZero to learn recursive neural programmer-interpreters. We empirically show that AlphaNPI-X can effectively learn to tackle challenging sparse manipulation tasks, such as stacking multiple blocks, where powerful model-free baselines fail.
- Open Access
Learning Compositional Neural Programs with Recursive Tree Search and Planning
We propose a novel reinforcement learning algorithm, AlphaNPI, that incorporates the strengths of Neural Programmer-Interpreters (NPI) and AlphaZero. NPI contributes structural biases in the form of modularity, hierarchy and recursion, which are helpful to reduce sample complexity, improve generalization and increase interpretability. AlphaZero contributes powerful neural network guided search algorithms, which we augment with recursion. AlphaNPI only assumes a hierarchical program specification with sparse rewards: 1 when the program execution satisfies the specification, and 0 otherwise. Using this specification, AlphaNPI is able to train NPI models effectively with RL for the first time, completely eliminating the need for strong supervision in the form of execution traces. The experiments show that AlphaNPI can sort as well as previous strongly supervised NPI variants. The AlphaNPI agent is also trained on a Tower of Hanoi puzzle with two disks and is shown to generalize to puzzles with an arbitrary number of disk
Open AccessLearning Compositional Neural Programs with Recursive Tree Search and Planning
We propose a novel reinforcement learning algorithm, AlphaNPI, that incorporates the strengths of Neural Programmer-Interpreters (NPI) and AlphaZero. NPI contributes structural biases in the form of modularity, hierarchy and recursion, which are helpful to reduce sample complexity, improve generalization and increase interpretability. AlphaZero contributes powerful neural network guided search algorithms, which we augment with recursion. AlphaNPI only assumes a hierarchical program specification with sparse rewards: 1 when the program execution satisfies the specification, and 0 otherwise. Using this specification, AlphaNPI is able to train NPI models effectively with RL for the first time, completely eliminating the need for strong supervision in the form of execution traces. The experiments show that AlphaNPI can sort as well as previous strongly supervised NPI variants. The AlphaNPI agent is also trained on a Tower of Hanoi puzzle with two disks and is shown to generalize to puzzles with an arbitrary number of disk