This thesis improves protein contact prediction using unsupervised and supervised methods.
problem Improving accuracy of protein contact prediction.
method Unsupervised and supervised deep learning methods.
result A scoring system called diversity score for measuring contact novelty.
Deep learning predicts protein contacts with high accuracy.
problem Low quality contact predictions for proteins without homologs.
method Integrates evolutionary coupling and sequence conservation through an ultra-deep neural network.
result Significantly outperforms existing methods in contact prediction and ab initio folding.
Motivation. Protein contact map describes the pairwise spatial and functional relationship of residues in a protein and contains key information for protein 3D structure prediction. Although studied extensively, it remains very challenging to predict contact map using only sequence information. Most existing methods pr…
Protein contacts contain important information for protein structure and functional study, but contact prediction from sequence remains very challenging. Both evolutionary coupling (EC) analysis and supervised machine learning methods are developed to predict contacts, making use of different types of information, resp…
mGPfusion predicts protein stability changes using a novel Gaussian process method.
problem Limited experimental data for predicting protein stability changes.
method Bayesian data fusion model combining experimental and molecular simulation data.
result mGPfusion outperforms state-of-the-art methods in predicting protein stability.
Study shows protein folding rate linked to topological changes.
problem Understanding protein folding kinetics and topology.
method Gauss linking integral, torsion, and sequence-distant contacts.
result Protein topology shifts from right-handed to left-handed with decreasing folding rate, associated with more sequence-distant contacts.
Symmetric CNNs improve sequential recommendation and protein structure prediction.
problem Improving prediction accuracy in sequential recommendation and protein structure inference.
method Developed a CNN architecture that preserves symmetry in convolutional layers, using parameterized convolutional kernels.
result Symmetric structured CNNs achieve better performance with fewer parameters.
Deep learning model predicts protein-ligand binding modes from docking data.
problem Improving protein-ligand binding mode prediction accuracy.
method Dual-graph architecture with separate sub-networks for ligand topology and protein-ligand interactions.
result Deep learning model outperforms docking programs in binding mode prediction.
New method maps protein sequences to embeddings encoding structural information.
problem Inferring structural properties from amino acid sequences when structures are unknown.
method Representation learning using bidirectional LSTM models with structural similarity and residue contact maps.
result Trained embeddings improve structural similarity prediction and transfer to other tasks.
A new method estimates protein evolutionary fields and couplings from alignments.
problem Estimating evolutionary fields and couplings from protein sequence alignments.
method Boltzmann machine with parallel, persistent Markov chain Monte Carlo method.
result Improved precision in predicting contact residue pairs.
Polynomial invariants classify molecular chains based on their contact arrangements.
problem No established invariants for molecular chains with both hard and soft contacts.
method Developed polynomial invariants for circuit topology of molecular chains.
result Polynomial invariants efficiently classify chains with various contact types.
Motivation: Prediction of ligands for proteins of known 3D structure is important to understand structure-function relationship, predict molecular function, or design new drugs. Results: We explore a new approach for ligand prediction in which binding pockets are represented by atom clouds. Each target pocket is compar…
A new model explains protein interactions via electron delocalization.
problem Understanding how protein interactions affect each other.
method Quantized discrete differential geometry of n-simplices.
result Allosteric regulation follows from the model of interactions.
Deep neural networks predict protein functions from sequences.
problem Accurately predicting protein functions from amino acid sequences.
method Artificial recurrent neural networks (RNN) with LSTM units trained on annotated datasets.
result RNN models achieved high performance for in-class and out-of-class protein function predictions.
PANDA predicts protein binding affinity changes from sequences, outperforming existing methods.
problem Accurately predicting changes in protein binding affinity due to mutations.
method Sequence-based machine learning approach using protein sequence information.
result PANDA achieves higher Pearson correlation coefficients than existing methods.
WideDTA predicts drug-target binding affinity using text-based information.
problem Predicting drug-target binding affinity is a major challenge in drug discovery.
method WideDTA uses chemical and biological textual sequence information, including protein sequence, ligand SMILES, protein domains and motifs, and maximum common substructure words.
result WideDTA outperformed DeepDTA on the KIBA dataset, indicating the word-based sequence representation is a promising alternative.
EBM predicts protein conformations at atomic scale using crystallized data.
problem Predicting the conformation of a side chain from its context within a protein structure.
method Energy-based model trained on crystallized protein data, evaluating performance on rotamer recovery task.
result EBM achieves performance close to state-of-the-art methods, including Rosetta energy function.
Tail-GNNs improve protein function prediction using relational reinforcement.
problem Predicting hierarchical protein functions from sequence data.
method Combining Tail-GNNs with dilated convolutional networks for multi-task learning.
result Significant improvement in F_1 score for protein function prediction.
Flexible Kernels for Protein Property Prediction
problem Predicting protein properties from sparse experimental data
method Sequence kernels using evolutionary substitution matrices and local linearity
result Data-efficient models of protein property landscapes
Deep learning predicts protein structures accurately.
problem Predicting the 3D structure of proteins from amino acid sequences.
method Embeddings and deep learning models for backbone atom distance matrices and torsion angles.
result Competitive results in CASP13 and CASP12, surpassing previous winners.
New method predicts protein structural quality using random forests.
problem Predicting protein structural quality from models.
method Combines multi/single model quality assessment with chemical, physical, geometric features and random forests.
result Random forest technique improves local quality assessment accuracy.
New multitask algorithm separates rare from frequent protein functions.
problem Challenging automated protein function prediction with unbalanced data.
method Uses dissimilarity information to separate rare class labels, unlike similarity-based approaches.
result Multitask label propagation algorithm performs best with dissimilarity matrix.
Sequence-based model predicts protein-protein interactions with high accuracy.
problem Predicting protein-protein interactions for alternative treatment options.
method Sequence clustering, discrete cosine transform, supervised machine learning, SVM with RBF.
result Mesh model achieved an average AUC of 0.84.
ProtTrans models predict protein features without evolutionary info.
problem Predicting protein features from amino acid sequences.
method Self-supervised deep learning on large protein datasets.
result ProtT5 embeddings outperform state-of-the-art for per-residue predictions.
Deep model learns protein interfaces from high-order interactions.
problem Predicting protein interfaces from amino acid pairs.
method Graph neural networks and convolutional neural networks for 2D dense predictions.
result Our method consistently improves interface prediction performance.
A new framework uses text descriptions to improve protein design.
problem Lack of effective methods to incorporate textual descriptions in protein design.
method ProteinDT framework that combines text and protein structural information.
result ProteinDT significantly improves protein design accuracy and performance.
ChemBoost predicts protein-ligand binding affinity using SMILES syntax.
problem Predicting high affinity drug-target interactions from sequence similarity alone.
method ChemBoost uses SMILES syntax to represent ligands as documents and proteins as sequences or ligand-centric features. It learns chemical word embeddings and predicts affinities using eXtreme Gradient Boosting.
result ChemBoost outperforms state-of-the-art systems in predicting protein-ligand affinities.
A new method predicts compounds for orphan proteins.
problem Predicting binding affinities for orphan proteins.
method Corresponding projections for transfer learning.
result The method outperforms state-of-the-art in orphan screening.
Two deep learning models predict protein-protein interactions with high accuracy.
problem Overfitting and information leak in deep learning models for PPI prediction.
method Carefully designed deep learning models, strict conditions for training and testing, and methodology to avoid information leak.
result Best model predicts more than 78% of human PPI with strong confidence.
Machine learning predicts protein structures and simulates dynamics.
problem Understanding and predicting protein folding and dynamics.
method Machine learning techniques for structure prediction and simulation.
result Machine learning enhances protein simulation and structure prediction.
AFP-CKSAAP predicts antifreeze proteins using k-spaced amino acid pairs with deep neural networks.
problem Predicting antifreeze proteins due to their diverse sequence characteristics.
method Deep neural network with skipped connections and ReLU non-linearity to learn protein sequence descriptors.
result AFP-CKSAAP achieves excellent prediction scores and high Youden's index (0.82) on independent dataset.
Study develops large margin machine learning models for predicting host-pathogen protein interactions.
problem Identifying host-pathogen interactions to develop new drugs for infectious diseases.
method Large margin machine learning models, specifically SVM with weighted negative sampling and distance-based weight assignment.
result Proposed and validated a new method for predicting host-pathogen protein interactions.
A new machine-learned CG model predicts protein structures efficiently.
problem Developing a universal, computationally efficient protein simulation model.
method Combining deep learning with all-atom protein simulations to create a transferable CG force field.
result The model predicts protein structures, intermediates, and fluctuations efficiently.
Continuous-depth Evoformer reduces protein folding prediction time and resource usage.
problem Efficient protein structure prediction with reduced computational costs.
method Continuous-depth formulation of Evoformer using Neural Ordinary Differential Equations (Neural ODEs).
result The continuous-time Evoformer achieves constant memory cost and improved efficiency.
TIP model improves POSE prediction with less resources.
problem Predicting polypharmacy side effects from drug-protein interactions.
method TIP model operates on three subgraphs for progressive representation learning.
result Improves accuracy by 7%+, time efficiency by 83imes, and space efficiency by 3imes. Article compares different machine learning techniques for protein classification.
problem Predicting enzyme class from unknown proteins is challenging.
method Implemented seven classification techniques on 4368 protein data.
result C5.0 classification technique gives highest accuracy and balanced performance.
Protein Thoughts interprets protein interactions with clear reasoning, improving prediction accuracy.
problem Lack of mechanistic justification in protein-protein interaction predictions.
method Interpretable search problem reformulation, hypothesis-guided entropy-regularized Tree-of-Thoughts search, embedding-space flow matching.
result Improves mean best-binder rank from 47.7 to 11.2 on SHS148k benchmark.
A new method predicts protein functions using variable-length sequences.
problem Computational methods for protein function prediction are slow and inaccurate for long sequences.
method Two feature sets: single fixed-sized segments and multi-sized segments, using bi-directional LSTM. Combined with MLDA features.
result Significant improvement in accuracy for long protein sequences.
Deep neural network improves amino acid side chain prediction accuracy.
problem Predicting amino acid side chain conformation for protein modeling and design.
method Deep neural network architecture without physics-based assumptions.
result Improved accuracy by more than 25% for aromatic residues.
PS8-Net improves eight-state protein secondary structure prediction accuracy.
problem Precise prediction of eight-state protein secondary structure (PSS) is crucial in bioinformatics.
method PS8-Net is a new deep convolutional neural network (DCNN) that uses a PS8 module with skip connections to enhance accuracy.
result PS8-Net achieves 76.89% Q8 accuracy on benchmark datasets.
New method predicts protein features using statistical relational learning.
problem Challenges in automatic protein feature annotation due to limited homology data.
method Introduces Semantically Based Regularization to incorporate prior knowledge.
result Improved overall prediction quality with constraints.
End-to-end model predicts protein interfaces from atomic coordinates.
problem Improving protein interface prediction using large datasets.
method Developed SASNet, an end-to-end learning model using only atomic coordinates.
result SASNet outperforms state-of-the-art methods trained on gold-standard data.
Machine learning predicts signaling peptides from protein star graphs.
problem Predicting signaling activity of proteins from molecular structure.
method Protein star graphs, S2SNet topological indices, Machine Learning (SVM-RFE, Laplacian kernel).
result Best model predicts 98.0% signaling pathways with AUROC 0.961.
DeepAffinity predicts compound-protein affinity from sequences, outperforming existing methods.
problem Lack of methods to predict compound-protein affinity from sequences alone.
method Unified RNN/GCNN-CNN model that unifies recurrent and convolutional neural networks.
result Model outperforms conventional options in predicting affinities with high accuracy.
CNN scoring function predicts protein-ligand interactions.
problem Scoring protein-ligand interactions for drug discovery.
method Convolutional Neural Networks (CNN) for 3D protein-ligand interactions.
result CNN scoring function outperforms AutoDock Vina in ranking poses.
Improved protein structure classification using weighted graphlets and deep neural networks.
problem Protein structure classification for function prediction.
method Developed a weighted network and graphlet-based measure, combined with a deep neural network.
result Significantly improved performance on 36 real datasets compared to existing methods.
MutaGAN predicts mutations of evolving protein populations using GANs.
problem Predicting mutations in evolving protein populations.
method Generative adversarial networks (GANs) with recurrent neural networks (RNNs).
result MutaGAN generates complete protein sequences with mutations.
ProteinNet provides a standardized data set for protein structure prediction.
problem Lack of standardized data sets for protein structure prediction.
method Created high-quality sequence alignments, multiple data splits, and validation sets.
result Facilitates fair assessment of machine learning models for protein structure.