Hierarchical text classification (HTC) is a complex subtask under multi-label text classification, characterized by a hierarchical label taxonomy and data imbalance. The best-performing models aim to learn a static representation by combining document and hierarchical label information. However, the relevance of document sections can vary based on the hierarchy level, necessitating a dynamic document representation. To address this, we propose HiGen, a text-generation-based framework utilizing language models to encode dynamic text representations. We introduce a level-guided loss function to capture the relationship between text and label name semantics. Our approach incorporates a task-specific pretraining strategy, adapting the language model to in-domain knowledge and significantly enhancing performance for classes with limited examples. Furthermore, we present a new and valuable dataset called ENZYME, designed for HTC, which comprises articles from PubMed with the goal of predicting Enzyme Commission (EC) numbers. Through extensive experiments on the ENZYME dataset and the widely recognized WOS and NYT datasets, our methodology demonstrates superior performance, surpassing existing approaches while efficiently handling data and mitigating class imbalance. We release our code and dataset here: https://github.com/viditjain99/HiGen.

MOTIVATION: Sphagnum-dominated peatlands store a substantial amount of terrestrial carbon. The genus is undersampled and under-studied. No experimental crystal structure from any Sphagnum species exists in the Protein Data Bank and fewer than 200 Sphagnum-related genes have structural models available in the AlphaFold Protein Structure Database. Tools and resources are needed to help bridge these gaps, and to enable the analysis of other structural proteomes now made possible by accurate structure prediction. RESULTS: We present the predicted structural proteome (25 134 primary transcripts) of Sphagnum divinum computed using AlphaFold, structural alignment results of all high-confidence models against an annotated nonredundant crystallographic database of over 90,000 structures, a structure-based classification of putative Enzyme Commission (EC) numbers across this proteome, and the computational method to perform this proteome-scale structure-based annotation. AVAILABILITY AND IMPLEMENTATION: All data and code are available in public repositories, detailed at https://github.com/BSDExabio/SAFA. The structural models of the S. divinum proteome have been deposited in the ModelArchive repository at https://modelarchive.org/doi/10.5452/ma-ornl-sphdiv.

MOTIVATION: From evolutionary interference, function annotation to structural prediction, protein sequence comparison has provided crucial biological insights. While many sequence alignment algorithms have been developed, existing approaches often cannot detect hidden structural relationships in the 'twilight zone' of low sequence identity. To address this critical problem, we introduce a computational algorithm that performs protein Sequence Alignments from deep-Learning of Structural Alignments (SAdLSA, silent 'd'). The key idea is to implicitly learn the protein folding code from many thousands of structural alignments using experimentally determined protein structures. RESULTS: To demonstrate that the folding code was learned, we first show that SAdLSA trained on pure α-helical proteins successfully recognizes pairs of structurally related pure β-sheet protein domains. Subsequent training and benchmarking on larger, highly challenging datasets show significant improvement over established approaches. For challenging cases, SAdLSA is ∼150% better than HHsearch for generating pairwise alignments and ∼50% better for identifying the proteins with the best alignments in a sequence library. The time complexity of SAdLSA is O(N) thanks to GPU acceleration. AVAILABILITY AND IMPLEMENTATION: Datasets and source codes of SAdLSA are available free of charge for academic users at http://sites.gatech.edu/cssb/sadlsa/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The problem of person re-identification, identifying the same person appeared in different camera views, is an important and challenging task in computer vision that has high potential application in areas like visual surveillance. In this paper we introduce a new feature fusion strategy for person reidentification that combines low-level Weighted Histograms of Overlapping Stripes (WHOS) features with mid-level color name descriptors and we adopt KISSME algorithm for person matching. Experiments on several public person reidentification datasets (VIPeR, i-LIDS and CAVIAR4REID) demonstrate that our approach achieves much better results compared with other state-of-the-art approaches.

MOTIVATION: Most proteins interact with small-molecule ligands such as metabolites or drug compounds. Over the past several decades, many of these interactions have been captured in high-resolution atomic structures. From a geometric point of view, most interaction sites for grasping these small-molecule ligands, as revealed in these structures, form concave shapes, or 'pockets', on the protein's surface. An efficient method for comparing these pockets could greatly assist the classification of ligand-binding sites, prediction of protein molecular function and design of novel drug compounds. RESULTS: We introduce a computational method, APoc (Alignment of Pockets), for the large-scale, sequence order-independent, structural comparison of protein pockets. A scoring function, the Pocket Similarity Score (PS-score), is derived to measure the level of similarity between pockets. Statistical models are used to estimate the significance of the PS-score based on millions of comparisons of randomly related pockets. APoc is a general robust method that may be applied to pockets identified by various approaches, such as ligand-binding sites as observed in experimental complex structures, or predicted pockets identified by a pocket-detection method. Finally, we curate large benchmark datasets to evaluate the performance of APoc and present interesting examples to demonstrate the usefulness of the method. We also demonstrate that APoc has better performance than the geometric hashing-based method SiteEngine. AVAILABILITY AND IMPLEMENTATION: The APoc software package including the source code is freely available at http://cssb.biology.gatech.edu/APoc.

Many biological activities originate from interactions between small-molecule ligands and their protein targets. A detailed structural and physico-chemical characterization of these interactions could significantly deepen our understanding of protein function and facilitate drug design. Here, we present a large-scale study on a non-redundant set of about 20,000 known ligand-binding sites, or pockets, of proteins. We find that the structural space of protein pockets is crowded, likely complete, and may be represented by about 1,000 pocket shapes. Correspondingly, the growth rate of novel pockets deposited in the Protein Data Bank has been decreasing steadily over the recent years. Moreover, many protein pockets are promiscuous and interact with ligands of diverse scaffolds. Conversely, many ligands are promiscuous and interact with structurally different pockets. Through a physico-chemical and structural analysis, we provide insights into understanding both pocket promiscuity and ligand promiscuity. Finally, we discuss the implications of our study for the prediction of protein-ligand interactions based on pocket comparison.

MOTIVATION: Protein-protein interactions play an essential role in many cellular processes. The rapid accumulation of protein-protein complex structures provides an unprecedented opportunity for comparative studies of protein-protein interactions. To facilitate such studies, it is necessary to develop an accurate and efficient computational algorithm for the comparison of protein-protein interaction modes. While there are many structural comparison approaches developed for individual proteins, very few methods are available for protein-protein complexes. RESULTS: We present a novel interface alignment method, iAlign, for the structural alignment of protein-protein interfaces. New scoring schemes for measuring interface similarity are introduced, and an iterative dynamic programming algorithm is implemented. We find that the similarity scores follow extreme value distributions. Using statistical models, we empirically estimate their statistical significance, which is in good agreement with manual classifications by human experts. Large-scale tests of iAlign were conducted on both artificial docking models and experimental structures. In a benchmark test on 1517 dimers, iAlign successfully detects biologically related, structurally similar protein-protein interfaces at a coverage percentage of 90% and an error per query of 0.05. When compared against previously published methods, iAlign is substantially more accurate and efficient. AVAILABILITY: The iAlign software package is freely available at http://cssb.biology.gatech.edu/iAlign.

UNLABELLED: In the post-genomic era, the annotation of protein function facilitates the understanding of various biological processes. To extend the range of function annotation methods to the twilight zone of sequence identity, we have developed approaches that exploit both protein tertiary structure and/or protein sequence evolutionary relationships. To serve the scientific community, we have integrated the structure prediction tools, TASSER, TASSER-Lite and METATASSER, and the functional inference tools, FINDSITE, a structure-based algorithm for binding site prediction, Gene Ontology molecular function inference and ligand screening, EFICAz(2), a sequence-based approach to enzyme function inference and DBD-hunter, an algorithm for predicting DNA-binding proteins and associated DNA-binding residues, into a unified web resource, Protein Structure and Function prediction Resource (PSiFR). AVAILABILITY AND IMPLEMENTATION: PSiFR is freely available for use on the web at http://psifr.cssb.biology.gatech.edu/

DNA-protein interactions are involved in many essential biological activities. Because there is no simple mapping code between DNA base pairs and protein amino acids, the prediction of DNA-protein interactions is a challenging problem. Here, we present a novel computational approach for predicting DNA-binding protein residues and DNA-protein interaction modes without knowing its specific DNA target sequence. Given the structure of a DNA-binding protein, the method first generates an ensemble of complex structures obtained by rigid-body docking with a nonspecific canonical B-DNA. Representative models are subsequently selected through clustering and ranking by their DNA-protein interfacial energy. Analysis of these encounter complex models suggests that the recognition sites for specific DNA binding are usually favorable interaction sites for the nonspecific DNA probe and that nonspecific DNA-protein interaction modes exhibit some similarity to specific DNA-protein binding modes. Although the method requires as input the knowledge that the protein binds DNA, in benchmark tests, it achieves better performance in identifying DNA-binding sites than three previously established methods, which are based on sophisticated machine-learning techniques. We further apply our method to protein structures predicted through modeling and demonstrate that our method performs satisfactorily on protein models whose root-mean-square Calpha deviation from native is up to 5 A from their native structures. This study provides valuable structural insights into how a specific DNA-binding protein interacts with a nonspecific DNA sequence. The similarity between the specific DNA-protein interaction mode and nonspecific interaction modes may reflect an important sampling step in search of its specific DNA targets by a DNA-binding protein.

Diverse mechanisms for DNA-protein recognition have been elucidated in numerous atomic complex structures from various protein families. These structural data provide an invaluable knowledge base not only for understanding DNA-protein interactions, but also for developing specialized methods that predict the DNA-binding function from protein structure. While such methods are useful, a major limitation is that they require an experimental structure of the target as input. To overcome this obstacle, we develop a threading-based method, DNA-Binding-Domain-Threader (DBD-Threader), for the prediction of DNA-binding domains and associated DNA-binding protein residues. Our method, which uses a template library composed of DNA-protein complex structures, requires only the target protein's sequence. In our approach, fold similarity and DNA-binding propensity are employed as two functional discriminating properties. In benchmark tests on 179 DNA-binding and 3,797 non-DNA-binding proteins, using templates whose sequence identity is less than 30% to the target, DBD-Threader achieves a sensitivity/precision of 56%/86%. This performance is considerably better than the standard sequence comparison method PSI-BLAST and is comparable to DBD-Hunter, which requires an experimental structure as input. Moreover, for over 70% of predicted DNA-binding domains, the backbone Root Mean Square Deviations (RMSDs) of the top-ranked structural models are within 6.5 A of their experimental structures, with their associated DNA-binding sites identified at satisfactory accuracy. Additionally, DBD-Threader correctly assigned the SCOP superfamily for most predicted domains. To demonstrate that DBD-Threader is useful for automatic function annotation on a large-scale, DBD-Threader was applied to 18,631 protein sequences from the human genome; 1,654 proteins are predicted to have DNA-binding function. Comparison with existing Gene Ontology (GO) annotations suggests that approximately 30% of our predictions are new. Finally, we present some interesting predictions in detail. In particular, it is estimated that approximately 20% of classic zinc finger domains play a functional role not related to direct DNA-binding.