Biology subjects

AlQuraishi, M.

Publications and source records attributed to AlQuraishi, M..

15 recordsLinked to original sources

bioRxiv2026

Quantitative machine learning of protein interactions reveals the multiscale organization and molecular syntax of signaling networks

Cells employ dense networks of transient protein-protein interactions mediated by modular peptide-binding domains and unstructured peptidic motifs for high-fidelity information processing. How these networks physically execute computations through protein interactions governed by complex intra- and intermolecular mechanisms remains indiscernible from current, sparse and non-quantitative, maps of the human interactome. Here, we introduce a quantitative statistical mechanical modeling (QSM) approach for machine learning domain-peptide affinities with experimental-level accuracy. Leveraging a new, principled algorithm for data harmonization and a biophysically informed neural network architecture, QSM learns to predict dissociation constants directly from amino acid sequences with calibrated confidence. We use QSM to construct the first quantitative drafts of human signaling networks and study these networks across three physical scales--recognition mechanisms of modular binding domains, combinatorial logic of multi-dentate proteins, and pathways inferred from de novo inference of protein interaction networks. We find that (i) modular domains, based on their binding preferences, selectivities, and strengths, fall into a limited number of biophysical equivalence groups, (ii) those domains, along with peptidic motifs, are "syntactically" combined within proteins to yield multivalent recognition mechanisms, and (iii) the organization of cellular function can be traced back to algorithmically detectable modules induced by domain-mediated interactions. In aggregate, these analyses instantiate a tractable roadmap towards a comprehensive, mechanistic, and simulatable articulation of the systems biology of signaling.

systems biology↗

bioRxiv2026

The first OpenBind release: An open experimental structure-affinity dataset and benchmark for structure-based AI

High-quality experimental datasets that link protein-ligand structures with binding affinity data are essential for developing and evaluating structure-based machine learning methods. To help address this need, we established OpenBind as an open-science initiative to generate large-scale experimental datasets for structure-based AI and molecular discovery. Here, we describe the first public OpenBind release, which, to the best of our knowledge, is the largest public single-target experimental structure-affinity dataset. The dataset focuses on enteroviral 2A protease, comprising 925 crystallographic binding events from 699 compounds and associated affinity measurements for 601 compounds. It combines structures from an initial fragment screen and follow-on molecules, together with affinity data, linking experimentally determined protein-ligand binding modes to biophysical measurements within a coherent antiviral discovery campaign. We used this dataset to evaluate protein-ligand structure prediction, binding-affinity prediction, and virtual screening using representative structure-based methods, including docking and cofolding. This exposed several challenges that are central to practical structure-based modelling: docking performance depends strongly on binding-pocket conformation, poses are difficult to rank, and structure-based affinity prediction remains challenging. Fine-tuning OpenFold3-p2 on the fragment-screen structures substantially improved pose prediction and virtual screening for related follow-on compounds, demonstrating how early-stage experimental structures can support target-specific model adaptation.

bioinformatics↗

bioRxiv2026

Learning the structural diversity in random protein sequence space

The universe of possible protein sequences is astronomically large, yet our understanding of the sequence-structure relationship is confined to the infinitesimal fraction used currently by life. Determining whether "foldable" architectures are rare singularities or accessible solutions is critical for understanding protein evolution and designing novel proteins. Here, we map the structural landscape of random sequence space by screening one million synthetic proteins using a high-throughput in vivo FRET biosensor. We reveal that this space is structurally heterogeneous, populated not only by disordered chains and stress-inducing aggregates but also by "benign" compact structures that resemble globular proteins and evade cellular chaperone responses. By training machine learning models on these phenotypes, we show that structural potential is learnable and generalizes to natural proteomes. These findings demonstrate that biology-like folds are accessible from random sequences with surprising frequency, providing data required to expand generative protein design beyond evolutionary priors.

synthetic biology↗

bioRxiv2026

Fast and Ultra-Capable Protein Design: Advancing the Frontier Through Atomistic SE(3)-Equivariance with Genie 3

Despite the breakneck pace of progress in protein design methodology, frontier problems remain challenging, with leading methods struggling to design high-affinity binders, scaffold multiple functional motifs, or stabilize large multi-domain proteins. Recent research efforts have focused on two areas: improving model reasoning when generating active sites or binding interfaces, and improving concordance between the design process and the in silico oracle used to select promising designs. In addressing the first, the field has shifted towards all-atom models that capture sidechain conformations in atomistic detail by eschewing data-efficient SE(3)-equivariance, mirroring the evolution of AlphaFold2 to AlphaFold3. In addressing the second, recent work has focused on replacing generative models employing diffusion or flow-matching with hallucination approaches that directly optimize the oracle in sequence space; this improves success rates but reduces computational efficiency. Here, we close and surpass the generation-hallucination gap by revisiting SE(3)-equivariance using a branched polymer treatment of protein structures. The resulting diffusion model, Genie 3, achieves state-of-the-art performance on binder design, motif scaffolding, and unconditional generation, while being significantly faster than the best existing methods. We use Genie 3 to design a nanomolar binder of Nipah Glycoprotein G, a tetramer with minimal structural or biophysical characterization, as part of the Adaptyv Bio Nipah Competition, achieving a 12.5% success rate. Taken together, our results present a new frontier in protein design capability and a reexamination of the role of SE(3)-equivariance in molecular modeling.

bioinformatics↗

bioRxiv2025

GREmLN: A Cellular Regulatory Network-Aware Transcriptomics Foundation Model

AO_SCPLOWBSTRACTC_SCPLOWThe ever-increasing availability of large-scale single-cell profiles presents an opportunity to develop foundation models to capture cell properties and behavior. However, standard language models such as transformers benefits from sequentially structured data with well defined absolute or relative positional relationships, while single cell RNA data have orderless gene features. Molecular-interaction graphs, such as gene regulatory networks (GRN) or protein-protein interaction (PPI) networks, offer graph structure-based models that effectively encode both non-local gene token dependencies, as well as potential causal relationships. We introduce GREmLN (Gene Regulatory Embedding-based Large Neural model), a foundation model that leverages graph signal processing to embed gene token graph structure directly within its attention mechanism, producing biologically informed single cell specific gene embeddings. Our model faithfully captures transcriptomics landscapes and achieves superior performance relative to state-of-the-art baselines on cell type annotation, graph structure understanding, and fine-tuned reverse perturbation prediction tasks. It offers a unified and interpretable framework for learning high-capacity foundational representations that capture complex, long-range regulatory dependencies from high-dimensional single-cell transcriptomic data. Moreover, the incorporation of graph-structured inductive biases enables more parameter-efficient architectures and accelerates training convergence.

bioinformatics↗

bioRxiv2025

Predicting specificity of TCR-pMHC interactions using machine learning and biophysical models

Understanding the mechanism of T-cell activation and T-cell receptor (TCR) discrimination of MHC-presented epitope peptides (pMHCs) remains an open problem. Machine learning (ML)-based prediction of TCR specificity has gained considerable recent attention. However, the capacity of current models to generalize to peptides unseen during training is currently unknown. Here, we use a proprietary cancer-patient data set that profiles TCR binding to novel regions of peptide space to show that peptide generalization remains an unsolved problem. Specifically, we show that while ML methods have demonstrable utility in predicting TCR specificity for known peptides, they fail to generalize to novel peptides. We also show that physics-based methods utilizing classical energy functions outperform ML methods when predicting TCR binding to novel peptides but underperform them on known peptides. In light of these observations, we develop a new ML model that leverages general knowledge acquired by protein foundation models to achieve better or comparable performance than either ML or biophysical methods on both in- and out-of-distribution TCR-pMHC specificity prediction. We furthermore analyze model performance as a function of distance of TCR sequence specificity between the training and test sets to quantitatively characterize the generalization potential of any given TCR-pMHC model. Our analysis sheds light on the status of modeling TCR-pMHC interactions and suggests new paths forward for continued method development and data acquisition.

immunology↗

bioRxiv2025

AlphaFold as a Prior: Experimental StructureDetermination Conditioned on a Pretrained Neural Network

Advances in machine learning have transformed structural biology, enabling swift and accurate prediction of protein structure from sequence. However, challenges persist in capturing sidechain packing, condition-dependent conformational dynamics, and biomolecular interactions, primarily due to scarcity of high-quality training data. Emerging techniques, including cryo-electron tomography (cryo-ET) and high-throughput crystallography, promise vast new sources of structural data, but translating experimental observations into mechanistically interpretable atomic models remains a key bottleneck. Here, we address these challenges by improving the efficiency of structural analysis through combining experimental measurements with a landmark protein structure prediction method - AlphaFold2. We present an augmentation of AlphaFold2, ROCKET, that refines its predictions using cryo-EM, cryo-ET, and X-ray crystallography data, and demonstrate that this approach captures biologically important structural variation that AlphaFold2 does not. By performing structure optimization in the space of coevolutionary embeddings, rather than Cartesian coordinates, ROCKET automates difficult modeling tasks, such as flips of functional loops and domain rearrangements, that are beyond the scope of current state-of-the-art methods and, in some instances, even manual human modeling. The ability to efficiently sample these barrier-crossing rearrangements unlocks a new horizon for scalable and automated model building. Crucially, ROCKET does not require retraining of AlphaFold2 and is readily adaptable to multimers, ligand-cofolding, and other data modalities. Conversely, our differentiable crystal-lographic and cryo-EM target functions are capable of augmenting other structure prediction methods. ROCKET thus provides an extensible framework for the integration of experimental observables with biomolecular machine learning.

biophysics↗

bioRxiv2025

From Mechanistic Interpretability to Mechanistic Biology: Training, Evaluating, and Interpreting Sparse Autoencoders on Protein Language Models

Protein language models (pLMs) are powerful predictors of protein structure and function, learning through unsupervised training on millions of protein sequences. pLMs are thought to capture common motifs in protein sequences, but the specifics of pLM features are not well understood. Identifying these features would not only shed light on how pLMs work, but potentially uncover novel protein biology--studying the model to study the biology. Motivated by this, we train sparse autoencoders (SAEs) on the residual stream of a pLM, ESM-2. By characterizing SAE features, we determine that pLMs use a combination of generic features and family-specific features to represent a protein. In addition, we demonstrate how known sequence determinants of properties such as thermostability and subcellular localization can be identified by linear probing of SAE features. For predictive features without known functional associations, we hypothesize their role in unknown mechanisms and provide visualization tools to aid their interpretation. Our study gives a better understanding of the limitations of pLMs, and demonstrates how SAE features can be used to help generate hypotheses for biological mechanisms. We release our code, model weights and feature visualizer.

bioinformatics↗

bioRxiv2022

OpenFold: Retraining AlphaFold2 yields new insights into its learning mechanisms and capacity for generalization

AlphaFold2 revolutionized structural biology with the ability to predict protein structures with exceptionally high accuracy. Its implementation, however, lacks the code and data required to train new models. These are necessary to (i) tackle new tasks, like protein-ligand complex structure prediction, (ii) investigate the process by which the model learns, which remains poorly understood, and (iii) assess the models generalization capacity to unseen regions of fold space. Here we report OpenFold, a fast, memory-efficient, and trainable implementation of AlphaFold2. We train OpenFold from scratch, fully matching the accuracy of AlphaFold2. Having established parity, we assess OpenFolds capacity to generalize across fold space by retraining it using carefully designed datasets. We find that OpenFold is remarkably robust at generalizing despite extreme reductions in training set size and diversity, including near-complete elisions of classes of secondary structure elements. By analyzing intermediate structures produced by OpenFold during training, we also gain surprising insights into the manner in which the model learns to fold proteins, discovering that spatial dimensions are learned sequentially. Taken together, our studies demonstrate the power and utility of OpenFold, which we believe will prove to be a crucial new resource for the protein modeling community.

bioinformatics↗

bioRxiv2022

High-throughput deep learning variant effect prediction with Sequence UNET

Understanding the consequences of protein coding mutations is important for many applications in biology and medicine. The vast number of possible mutations across species makes comprehensive experimental characterisation impossible, even with recent high-throughput techniques, which means computationally predicting the consequences of variation is essential for many analyses. Previous variant effect prediction (VEP) tools, generally based on evolutionary conservation and protein structure, are often computationally intensive, making them difficult to scale and limiting potential applications. Recent developments in deep learning techniques, including protein language models, and biological data scale have led to a new generation of predictors. These models have improved prediction performance but are still often intensive to run because of slow training steps, hardware requirements and large model sizes. In this work we introduce a new highly scalable deep learning architecture, Sequence UNET, that classifies and predicts variant frequency directly from protein sequence. This model learns to build representations of protein sequence features at a range of scales using a fully convolutional U-shaped compression/expansion architecture. We show that it can generalise to pathogenicity prediction, achieving comparable performance on ClinVar to methods including EVE and ESM-1b at greatly reduced computational cost. We further demonstrate its scalability by analysing the consequences of 8.3 billion variants in 904,134 proteins detected in a large-scale proteomics analysis, showing a link between conservation and protein abundance. Sequence UNET can be run on modest hardware through an easy to use Python package.

bioinformatics↗

bioRxiv2021

Protein Structure Prediction Using a Maximum Likelihood Formulation of a Recurrent Geometric Network

Only [~]40% of the human proteome has structural coordinates available from experiment (i.e., X-ray crystallography, NMR spectroscopy, or cryo-EM) or homology modeling with quality templates (i.e., 30% sequence identity or greater), leaving most of the proteome structurally unsolved. Deep learning (DL) methods for predicting protein structure can help close knowledge gaps where experimental and homology models are difficult to obtain. Recent advances in these DL methods have shown promising results in expanding structural coverage to the scale of the entire human proteome, providing researchers with more complete protein structural information. Here, we improve upon an existing DL algorithm for protein structure prediction, the Recurrent Geometric Network (RGN). We first expand the training dataset to include experimental uncertainty data in the form of atomic displacement parameters, then derive a maximum likelihood loss function that incorporates this uncertainty data into model training. Compared to the original RGN, our novel maximum likelihood model improves the rate of convergence of initial model training and ultimately results in more accurate structure prediction according to the root mean square deviation (RMSD) of backbone atoms, the Global Distance Test (GDT), the Global Distance Test High Accuracy (GDT-HA), and the Template-Modeling Score (TM-Score). Our model also predicts structures with more favorable backbone torsions, which provide more accurate starting coordinates for downstream physics-based simulations. Based on these results, our maximum likelihood reformulation provides a framework for improving existing or future machine learning algorithms for protein structure prediction. The augmented dataset, data collection scripts, reformulated RGN source code, and a series of trained models are publicly available at https://github.com/SchniedersLab/likelihood-rgn.

biophysics↗

bioRxiv2020

Recombination and convergent evolution led to the emergence of 2019 Wuhan coronavirus

The emergence of SARS-CoV-2 underscores the need to better understand the evolutionary processes that drive the emergence and adaptation of zoonotic viruses in humans. In the betacoronavirus genus, which also includes SARS-CoV and MERS-CoV, recombination frequently encompasses the Receptor Binding Domain (RBD) of the Spike protein, which, in turn, is responsible for viral binding to host cell receptors. Here, we find evidence of a recombination event in the RBD involving ancestral linages to both SARS-CoV and SARS-CoV-2. Although we cannot specify the recombinant nor the parental strains, likely due to the ancestry of the event and potential undersampling, our statistical analyses in the space of phylogenetic trees support such an ancestral recombination. Consequently, SARS-CoV and SARS-CoV-2 share an RBD sequence that includes two insertions (positions 432-436 and 460-472), as well as the variants 427N and 436Y. Both 427N and 436Y belong to a helix that interacts directly with the human ACE2 (hACE2) receptor. Reconstruction of ancestral states, combined with protein-binding affinity analyses using the physics-based trRosetta algorithm, reveal that the recombination event involving ancestral strains of SARS-CoV and SARS-CoV-2 led to an increased affinity for hACE2 binding, and that alleles 427N and 436Y significantly enhanced affinity as well. Structural modeling indicates that ancestors of SARS-CoV-2 may have acquired the ability to infect humans decades ago. The binding affinity with the human receptor was subsequently boosted in SARS-CoV and SARS-CoV-2 through further mutations in RBD. In sum, we report an ancestral recombination event affecting the RBD of both SARS-CoV and SARS-CoV-2 that was associated with an increased binding affinity to hACE2. ImportanceThis paper addresses critical questions about the origin of the SARS-CoV-2 virus: what are the evolutionary mechanisms that led to the emergence of the virus, and how can we leverage such knowledge to assess the potential of SARS-like viruses to become pandemic strains? In this work, we demonstrate common mechanisms involved in the emergence of human-infecting SARS-like viruses: first, by acquiring a common haplotype in the RBD through recombination, and further, through increased specificity to the human ACE2 receptor through lineage specific mutations. We also show that the ancestors of SARS-CoV-2 already had the potential to infect humans at least a decade ago, suggesting that SARS-like viruses currently circulating in wild animal species constitute a source of potential pandemic re-emergence.

evolutionary biology↗

bioRxiv2019

Unified rational protein engineering with sequence-only deep representation learning

Rational protein engineering requires a holistic understanding of protein function. Here, we apply deep learning to unlabelled amino acid sequences to distill the fundamental features of a protein into a statistical representation that is semantically rich and structurally, evolutionarily, and biophysically grounded. We show that the simplest models built on top of this unified representation (UniRep) are broadly applicable and generalize to unseen regions of sequence space. Our data-driven approach reaches near state-of-the-art or superior performance predicting stability of natural and de novo designed proteins as well as quantitative function of molecularly diverse mutants. UniRep further enables two orders of magnitude cost savings in a protein engineering task. We conclude UniRep is a versatile protein summary that can be applied across protein engineering informatics.

synthetic biology↗

bioRxiv2018

pNeRF: Parallelized Conversion from Internal to Cartesian Coordinates

The conversion of polymer parameterization from internal coordinates (bond lengths, angles, and torsions) to Cartesian coordinates is a fundamental task in molecular modeling, often performed using the Natural Extension Reference Frame (NeRF) algorithm. NeRF can be parallelized to process multiple polymers simultaneously, but is not parallelizable along the length of a single polymer. A mathematically equivalent algorithm, pNeRF, has been derived that is parallelizable along a polymers length. Empirical analysis demonstrates an order-of-magnitude speed up using modern GPUs and CPUs. In machine learning-based workflows, in which partial derivatives are backpropagated through NeRF equations and neural network primitives, switching to pNeRF can reduce the fractional computational cost of coordinate conversion from over two-thirds to around 10%. An optimized TensorFlow-based implementation of pNeRF is available on GitHub.

bioinformatics↗

bioRxiv2018

End-to-end differentiable learning of protein structure

Predicting protein structure from sequence is a central challenge of biochemistry. Co-evolution methods show promise, but an explicit sequence-to-structure map remains elusive. Advances in deep learning that replace complex, human-designed pipelines with differentiable models optimized end-to-end suggest the potential benefits of similarly reformulating structure prediction. Here we report the first end-to-end differentiable model of protein structure. The model couples local and global protein structure via geometric units that optimize global geometry without violating local covalent chemistry. We test our model using two challenging tasks: predicting novel folds without co-evolutionary data and predicting known folds without structural templates. In the first task the model achieves state-of-the-art accuracy and in the second it comes within 1-2[A]; competing methods using co-evolution and experimental templates have been refined over many years and it is likely that the differentiable approach has substantial room for further improvement, with applications ranging from drug discovery to protein design.

bioinformatics↗