Unsupervised protein embeddings outperform hand-crafted sequence and structure features at predicting molecular function.
Ontology highlight
ABSTRACT: Motivation
Protein function prediction is a difficult bioinformatics problem. Many recent methods use deep neural networks to learn complex sequence representations and predict function from these. Deep supervised models require a lot of labeled training data which are not available for this task. However, a very large amount of protein sequences without functional labels is available.Results
We applied an existing deep sequence model that had been pretrained in an unsupervised setting on the supervised task of protein molecular function prediction. We found that this complex feature representation is effective for this task, outperforming hand-crafted features such as one-hot encoding of amino acids, k-mer counts, secondary structure and backbone angles. Also, it partly negates the need for complex prediction models, as a two-layer perceptron was enough to achieve competitive performance in the third Critical Assessment of Functional Annotation benchmark. We also show that combining this sequence representation with protein 3D structure information does not lead to performance improvement, hinting that 3D structure is also potentially learned during the unsupervised pretraining.Availability and implementation
Implementations of all used models can be found at https://github.com/stamakro/GCN-for-Structure-and-Function.Supplementary information
Supplementary data are available at Bioinformatics online.
SUBMITTER: Villegas-Morcillo A
PROVIDER: S-EPMC8055213 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:User Identity Linkage (UIL) is the problem of matching user identities across multiple online social networks (OSNs) which belong to the same person. The solutions to UIL problem facilitate cross-platform research on OSN users and enable many useful applications such as user profiling and recommendation. As the UIL labeled data are often lacking and costly to obtain, learning user embeddings for matching user identities using an unsupervised approach is therefore highly desired. In this paper, we propose a novel unsupervised UIL framework for enhancing existing user embedding-based UIL methods. Our proposed framework incorporates two key ideas, user-discriminative features and retrofitting embedding. The user-discriminative features enable us to differentiate a specific user identity from other users in its OSN. From the user-discriminative features, we derive pairs of similar user identities across OSNs for retrofitting the base user embeddings of existing UIL methods. Through extensive experiments on three real-world OSN datasets, we show that our framework can leverage user-discriminative features to improve the accuracy of different user embedding-based UIL methods significantly. The quantum of improvement can also be surprisingly good even for existing UIL methods with very poor matching accuracy. Electronic supplementary material The online version of this chapter (10.1007/978-3-030-47426-3_30) contains supplementary material, which is available to authorized users.
| S-EPMC7206306 | biostudies-literature
Project description:Objectives: To develop and validate the model for distinguishing brain abscess from cystic glioma by combining deep transfer learning (DTL) features and hand-crafted radiomics (HCR) features in conventional T1-weighted imaging (T1WI) and T2-weighted imaging (T2WI). Methods: This single-center retrospective analysis involved 188 patients with pathologically proven brain abscess (102) or cystic glioma (86). One thousand DTL and 105 HCR features were extracted from the T1WI and T2WI of the patients. Three feature selection methods and four classifiers, such as k-nearest neighbors (KNN), random forest classifier (RFC), logistic regression (LR), and support vector machine (SVM), for distinguishing brain abscess from cystic glioma were compared. The best feature combination and classifier were chosen according to the quantitative metrics including area under the curve (AUC), Youden Index, and accuracy. Results: In most cases, deep learning-based radiomics (DLR) features, i.e., DTL features combined with HCR features, contributed to a higher accuracy than HCR and DTL features alone for distinguishing brain abscesses from cystic gliomas. The AUC values of the model established, based on the DLR features in T2WI, were 0.86 (95% CI: 0.81, 0.91) in the training cohort and 0.85 (95% CI: 0.75, 0.95) in the test cohort, respectively. Conclusions: The model established with the DLR features can distinguish brain abscess from cystic glioma efficiently, providing a useful, inexpensive, convenient, and non-invasive method for differential diagnosis. This is the first time that conventional MRI radiomics is applied to identify these diseases. Also, the combination of HCR and DTL features can lead to get impressive performance.
| S-EPMC8636043 | biostudies-literature
Project description:The rapid expansion of protein sequence and structure databases has resulted in a significant number of proteins with ambiguous or unknown function. While advances in machine learning techniques hold great potential to fill this annotation gap, current methods for function prediction are unable to associate global function reliably to the specific residues responsible for that function. We address this issue by introducing PARSE (Protein Annotation by Residue-Specific Enrichment), a knowledge-based method which combines pre-trained embeddings of local structural environments with traditional statistical techniques to identify enriched functions with residue-level explainability. For the task of predicting the catalytic function of enzymes, PARSE achieves comparable or superior global performance to state-of-the-art machine learning methods (F1 score > 85%) while simultaneously annotating the specific residues involved in each function with much greater precision. Since it does not require supervised training, our method can make one-shot predictions for very rare functions and is not limited to a particular type of functional label (e.g. Enzyme Commission numbers or Gene Ontology codes). Finally, we leverage the AlphaFold Structure Database to perform functional annotation at a proteome scale. By applying PARSE to the dark proteome-predicted structures which cannot be classified into known structural families-we predict several novel bacterial metalloproteases. Each of these proteins shares a strongly conserved catalytic site despite highly divergent sequences and global folds, illustrating the value of local structure representations for new function discovery.
| S-EPMC10614799 | biostudies-literature
Project description:We have noted consistent structural similarities among unrelated proteases. In comparison with other proteins of similar size, proteases have smaller than average surface areas, smaller radii of gyration, and higher C(alpha) densities. These findings imply that proteases are, as a group, more tightly packed than other proteins. There are also notable differences in secondary structure content between these two groups of proteins: proteases have fewer helices and more loops. We speculate that both high packing density and low alpha-helical content coevolved in proteases to avoid autolysis. By using the structural parameters that seem to show some separation between proteases and nonproteases, a neural network has been trained to predict protease function with over 86% accuracy. Moreover, it is possible to identify proteases whose folds were not represented during training. Similar structural analyses may be useful for identifying other classes of proteins and may be of great utility for categorizing the flood of structures soon to flow from structural genomics initiatives.
| S-EPMC18123 | biostudies-literature
Project description:We present a new general-purpose machine learning model that is able to predict a variety of crystal properties, including Fermi level energy and band gap, as well as spectral ones such as electronic densities of states. The model is based on atomic representations that enable it to effectively capture complex information about each atom and its surrounding environment in a crystal. The accuracy achieved for band gaps exceeds results previously published. By design, our model is not restricted to the electronic properties discussed here but can be extended to fit diverse chemical descriptors. Its advantages are (a) its low computational requirements, making it an efficient tool for high-throughput screening of materials; and (b) the simplicity and flexibility of its architecture, facilitating implementation and interpretation, especially for researchers in the field of computational chemistry.
| S-EPMC11323004 | biostudies-literature
Project description:Early prediction of the potential for neurological recovery after resuscitation from cardiac arrest is difficult but important. Currently, no clinical finding or combination of findings are sufficient to accurately predict or preclude favorable recovery of comatose patients in the first 24 to 48 hours after resuscitation. Thus, life-sustaining therapy is often continued for several days in patients whose irrecoverable injury is not yet recognized. Conversely, early withdrawal of life-sustaining therapy increases mortality among patients who otherwise might have gone on to recover. In this work, we present Canonical Autocorrelation Analysis (CAA) and Canonical Autocorrelation Embeddings (CAE), novel methods suitable for identifying complex patterns in high-resolution multivariate data often collected in highly monitored clinical environments such as intensive care units. CAE embeds sets of datapoints onto a space that characterizes their latent correlation structures and allows direct comparison of these structures through the use of a distance metric. The methodology may be particularly suitable when the unit of analysis is not just an individual datapoint but a dataset, as for instance in patients for whom physiological measures are recorded over time, and where changes of correlation patterns in these datasets are informative for the task at hand. We present a proof of concept to illustrate the potential utility of CAE by applying it to characterize electroencephalographic recordings from 80 comatose survivors of cardiac arrest, aiming to identify patients who will survive to hospital discharge with favorable functional recovery. Our results show that with very low probability of making a Type 1 error, we are able to identify 32.5% of patients who are likely to have a good neurological outcome, some of whom have otherwise unfavorable clinical characteristics. Importantly, some of these had 5% predicted chance of favorable recovery based on initial illness severity measures alone. Providing this information to support clinical decision-making could motivate the continuation of life-sustaining therapies for these patients.
| S-EPMC6349311 | biostudies-literature
Project description:PurposeTo test the ability of machine learning classifiers (MLCs) using optical coherence tomography (OCT) and standard automated perimetry (SAP) parameters to discriminate between healthy and glaucomatous individuals, and to compare it to the diagnostic ability of the combined structure-function index (CSFI), general ophthalmologists and glaucoma specialists.DesignCross-sectional prospective study.MethodsFifty eight eyes of 58 patients with early to moderate glaucoma (median value of the mean deviation = -3.44 dB; interquartile range, -6.0 to -2.4 dB) and 66 eyes of 66 healthy individuals underwent OCT and SAP tests. The diagnostic accuracy (area under the ROC curve-AUC) of 10 MLCs was compared to those obtained with the CSFI, 3 general ophthalmologists and 3 glaucoma specialists exposed to the same OCT and SAP data.ResultsThe AUCs obtained with MLCs ranged from 0.805 (Classification Tree) to 0.931 (Radial Basis Function Network, RBF). The sensitivity at 90% specificity ranged from 51.6% (Classification Tree) to 82.8% (Bagging, Multilayer Perceptron and Support Vector Machine Gaussian). The CSFI had a sensitivity of 79.3% at 90% specificity, and the highest AUC (0.948). General ophthalmologists and glaucoma specialists' grading had sensitivities of 66.2% and 83.8% at 90% specificity, and AUCs of 0.879 and 0.921, respectively. RBF (the best MLC), the CSFI, and glaucoma specialists showed significantly higher AUCs than that obtained by general ophthalmologists (P<0.05). However, there were no significant differences between the AUCs obtained by RBF, the CSFI, and glaucoma specialists (P>0.25).ConclusionOur findings suggest that both MLCs and the CSFI can be helpful in clinical practice and effectively improve glaucoma diagnosis in the primary eye care setting, when there is no glaucoma specialist available.
| S-EPMC6281287 | biostudies-literature
Project description:ObjectiveTo evaluate the value of an integrated model incorporating deep learning (DL), hand-crafted radiomics and clinical and US imaging features for diagnosing central lymph node metastasis (CLNM) in patients with papillary thyroid cancer (PTC).MethodsThis retrospective study reviewed 613 patients with clinicopathologically confirmed PTC from two institutions. The DL model and hand-crafted radiomics model were developed using primary lesion images and then integrated with clinical and US features selected by multivariate analysis to generate an integrated model. The performance was compared with junior and senior radiologists on the independent test set. SHapley Additive exPlanations (SHAP) plot and Gradient-weighted Class Activation Mapping (Grad-CAM) were used for the visualized explanation of the model.ResultsThe integrated model yielded the best performance with an AUC of 0.841. surpassing that of the hand-crafted radiomics model (0.706, p < 0.001) and the DL model (0.819, p = 0.26). Compared to junior and senior radiologists, the integrated model reduced the missed CLNM rate from 57.89% and 44.74-27.63%, and decreased the rate of unnecessary central lymph node dissection (CLND) from 29.87% and 27.27-18.18%, respectively. SHAP analysis revealed that the DL features played a primary role in the diagnosis of CLNM, while clinical and US features (such as extrathyroidal extension, tumour size, age, gender, and multifocality) provided additional support. Grad-CAM indicated that the model exhibited a stronger focus on thyroid capsule in patients with CLNM.ConclusionIntegrated model can effectively decrease the incidence of missed CLNM and unnecessary CLND. The application of the integrated model can help improve the acceptance of AI-assisted US diagnosis among radiologists.
| S-EPMC10787418 | biostudies-literature
Project description:A large number of distal enhancers and proximal promoters form enhancer-promoter interactions to regulate target genes in the human genome. Although recent high-throughput genome-wide mapping approaches have allowed us to more comprehensively recognize potential enhancer-promoter interactions, it is still largely unknown whether sequence-based features alone are sufficient to predict such interactions.Here, we develop a new computational method (named PEP) to predict enhancer-promoter interactions based on sequence-based features only, when the locations of putative enhancers and promoters in a particular cell type are given. The two modules in PEP (PEP-Motif and PEP-Word) use different but complementary feature extraction strategies to exploit sequence-based information. The results across six different cell types demonstrate that our method is effective in predicting enhancer-promoter interactions as compared to the state-of-the-art methods that use functional genomic signals. Our work demonstrates that sequence-based features alone can reliably predict enhancer-promoter interactions genome-wide, which could potentially facilitate the discovery of important sequence determinants for long-range gene regulation.The source code of PEP is available at: https://github.com/ma-compbio/PEP .jianma@cs.cmu.edu.Supplementary data are available at Bioinformatics online.
| S-EPMC5870728 | biostudies-literature
Project description:Similarity-based search of sequence collections is a core task in bioinformatics, one dominated for most of the genomic era by exact and heuristic alignment-based algorithms. However, even efficient heuristics such as BLAST may not scale to the data sets now emerging, motivating a range of alignment-free alternatives exploiting the underlying lexical structure of each sequence. In this paper, we introduce two supervised approaches-SuperVec and SuperVecX-to learn sequence embeddings. These methods extend earlier Representation Learning (RepL) based methods to include class-related information for each sequence during training. Including class information ensures that related sequence fragments have proximal representations in the target space, better reflecting the structure of the domain. We show the quality of the embeddings learned through these methods on (i) sequence retrieval and (ii) classification tasks. We also propose an hierarchical tree-based approach specifically designed for the sequence retrieval problem. The resulting methods, which we term H-SuperVec or H-SuperVecX, according to their respective use of SuperVec or SuperVecX, learn embeddings across a range of feature spaces based on exclusive and exhaustive subsets of the class labels. Experiments show that the proposed methods perform better for retrieval and classification tasks over existing (unsupervised) RepL-based approaches. Further, the new methods are an order of magnitude faster than BLAST for the database retrieval task, supporting hybrid approaches that rapidly filter the collection so that only potentially relevant records remain. Such filtering of the original database allows slower but more accurate methods to be executed quickly over a far smaller dataset. Thus, we may achieve faster query processing and higher precision than before.
| S-EPMC7069636 | biostudies-literature