Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “protein representation learning”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5Linked to original sources
PubMed2006

Discovering semantic features in the literature: a foundation for building functional associations.

BACKGROUND: Experimental techniques such as DNA microarray, serial analysis of gene expression (SAGE) and mass spectrometry proteomics, among others, are generating large amounts of data related to genes and proteins at different levels. As in any other experimental approach, it is necessary to analyze these data in the context of previously known information about the biological entities under study. The literature is a particularly valuable source of information for experiment validation and interpretation. Therefore, the development of automated text mining tools to assist in such interpretation is one of the main challenges in current bioinformatics research. RESULTS: We present a method to create literature profiles for large sets of genes or proteins based on common semantic features extracted from a corpus of relevant documents. These profiles can be used to establish pair-wise similarities among genes, utilized in gene/protein classification or can be even combined with experimental measurements. Semantic features can be used by researchers to facilitate the understanding of the commonalities indicated by experimental results. Our approach is based on non-negative matrix factorization (NMF), a machine-learning algorithm for data analysis, capable of identifying local patterns that characterize a subset of the data. The literature is thus used to establish putative relationships among subsets of genes or proteins and to provide coherent justification for this clustering into subsets. We demonstrate the utility of the method by applying it to two independent and vastly different sets of genes. CONCLUSION: The presented method can create literature profiles from documents relevant to sets of genes. The representation of genes as additive linear combinations of semantic features allows for the exploration of functional associations as well as for clustering, suggesting a valuable methodology for the validation and interpretation of high-throughput experimental data.

Abstracting and Indexing↗
PubMed2002

Memory for context is impaired by injecting anisomycin into dorsal hippocampus following context exploration.

Pre-exposure to the context facilitates the small amount of contextual fear conditioning that is normally produced by immediate shock. This context pre-exposure facilitation effect provides a convenient way to study the rat's learning about context. We recently reported that anterograde damage to dorsal hippocampus prevents this facilitation. The present experiments strengthen this conclusion by showing that the protein synthesis inhibitor, anisomycin, injected bilaterally into the dorsal hippocampus following context pre-exposure also significantly reduces the facilitation effect. The same treatment given immediately after immediate shock, however, had no effect on facilitation. These results support theories that assume that, (a) contextual fear involves two processes, acquiring and storing a conjunctive representation of a context and associating that representation with fear; and (b) the hippocampus contributes to contextual fear by participating in the storage of the memory representation of the context.

Animals↗
PubMed1993

A role for the neural cell adhesion molecule in a late, consolidating phase of glycoprotein synthesis six hours following passive avoidance training of the young chick.

We have investigated the effect of intracranial injections of the amnestic anti-metabolite, 2-deoxygalactose, and antibodies to the neural cell adhesion molecule on retention of a one-trial passive avoidance task in chicks. Groups of chicks received bilateral intracranial injections of 10 mumol/hemisphere 2-deoxygalactose or 10 microliters/hemisphere anti-neural cell adhesion molecule and were tested 24 h following training. 2-Deoxygalactose injections were amnestic when administered at a previously established time (30 min pre-training). Here we show that the agent is also amnestic when injected within a second time window occurring specifically 6-8 h after training. Administration of 2-deoxygalactose between 2 and 6 h or after 8 h post-training was without effect on retention tested 24 h following training. Anti-neural cell adhesion molecule injections were amnestic only when performed at a time which coincided with the second phase of 2-deoxygalactose susceptibility. Further experiments demonstrated that the neural cell adhesion molecule is one of the molecules into which 2-deoxygalactose is incorporated. Additionally, we investigated the extent of diffusion of 2-deoxygalactose and anti-neural cell adhesion molecule following their injection, with respect to their residence in forebrain loci known to be involved in the memory for passive avoidance. We interpret these data as indicating that two waves of glycoprotein synthesis are necessary for the establishment of long-term memory for the experience of passive avoidance training. The evidence is discussed in the context of earlier results indicating that the two waves involve different glycoprotein species and, possibly, different forebrain regions. We speculate that the late phase of glycoprotein synthesis coincides with, and is required for, modulation of cell-cell adhesion processes, reflecting the selection and stabilization of synapses which maintain an enduring representation of long-term memory.

Analysis of Variance↗
PubMed2005

A refined model of sleep and the time course of memory formation.

Research in the neurosciences continues to provide evidence that sleep plays a role in the processes of learning and memory. There is less of a consensus, however, regarding the precise stages of memory development during which sleep is considered a requirement, simply favorable, or not important. This article begins with an overview of recent studies regarding sleep and learning, predominantly in the procedural memory domain, and is measured against our current understanding of the mechanisms that govern memory formation. Based on these considerations, I offer a new neurocognitive framework of procedural learning, consisting first of acquisition, followed by two specific stages of consolidation, one involving a process of stabilization, the other involving enhancement, whereby delayed learning occurs. Psychophysiological evidence indicates that initial acquisition does not rely fundamentally on sleep. This also appears to be true for the stabilization phase of consolidation, with durable representations, resistant to interference, clearly developing in a successful manner during time awake (or just time, per se). In contrast, the consolidation stage, resulting in additional/enhanced learning in the absence of further rehearsal, does appear to rely on the process of sleep, with evidence for specific sleep-stage dependencies across the procedural domain. Evaluations at a molecular, cellular, and systems level currently offer several sleep specific candidates that could play a role in sleep-dependent learning. These include the upregulation of select plasticity-associated genes, increased protein synthesis, changes in neurotransmitter concentration, and specific electrical events in neuronal networks that modulate synaptic potentiation.

Adaptation, Physiological↗
PubMed2000

Postnatal intracerebroventricular administrations of NGF alter spatial memory in adulthood.

The present work assessed the effects of intracerebroventricular injections (2x5 mg/2.5 ml) of recombined human nerve growth factor (rhNGF) at postnatal days 2 and 3 upon the development of spatial learning capacities in rats. The treated rats were trained at the age of 22 days to escape onto an invisible platform at a fixed position in space in a Morris navigation task. For half of the subjects, the training position was also cued, a procedure aimed at facilitating escape and reducing attention to the distant spatial cues. At the age of 2 months all the rats were retrained in the same task. Treatment effects were found in both immature and adult rats. The injection of NGF induced a slight alteration of the immature rats' performance. In contrast, a marked impairment of spatial abilities was shown in the 2-month-old rats. The most consistent effects were a significant increase in the escape latency and a decrease bias towards the training platform area during probe trials. The reduction of spatial memory was particularly marked if the subjects had been trained in a cued condition. Taken together, these experiments reveal that an acute pharmacological treatment that leads to transient modifications during early development might induce a behavioural change long after treatment. Thus, the development and the maintenance of an accurate spatial representation are tightly related to the development of brain structures that could be altered by precocious NGF administrations.

Aging↗
PubMed2003

Machine learning approaches to lung cancer prediction from mass spectra.

We addressed the problem of discriminating between 24 diseased and 17 healthy specimens on the basis of protein mass spectra. To prepare the data, we performed mass to charge ratio (m/z) normalization, baseline elimination, and conversion of absolute peak height measures to height ratios. After preprocessing, the major difficulty encountered was the extremely large number of variables (1676 m/z values) versus the number of examples (41). Dimensionality reduction was treated as an integral part of the classification process; variable selection was coupled with model construction in a single ten-fold cross-validation loop. We explored different experimental setups involving two peak height representations, two variable selection methods, and six induction algorithms, all on both the original 1676-mass data set and on a prescreened 124-mass data set. Highest predictive accuracies (1-2 off-sample misclassifications) were achieved by a multilayer perceptron and Naïve Bayes, with the latter displaying more consistent performance (hence greater reliability) over varying experimental conditions. We attempted to identify the most discriminant peaks (proteins) on the basis of scores assigned by the two variable selection methods and by neural network based sensitivity analysis. These three scoring schemes consistently ranked four peaks as the most relevant discriminators: 11683, 1403, 17350 and 66107.

Algorithms↗
PubMed2004

Use of a structural alphabet for analysis of short loops connecting repetitive structures.

BACKGROUND: Because loops connect regular secondary structures, analysis of the former depends directly on the definition of the latter. The numerous assignment methods, however, can offer different definitions. In a previous study, we defined a structural alphabet composed of 16 average protein fragments, which we called Protein Blocks (PBs). They allow an accurate description of every region of 3D protein backbones and have been used in local structure prediction. In the present study, we use this structural alphabet to analyze and predict the loops connecting two repetitive structures. RESULTS: We first analyzed the secondary structure assignments. Use of five different assignment methods (DSSP, DEFINE, PCURVE, STRIDE and PSEA) showed the absence of consensus: 20% of the residues were assigned to different states. The discrepancies were particularly important at the extremities of the repetitive structures. We used PBs to describe and predict the short loops because they can help analyze and in part explain these discrepancies. An analysis of the PB distribution in these regions showed some specificities in the sequence-structure relationship. Of the amino acid over- or under-representations observed in the short loop databank, 20% did not appear in the entire databank. Finally, predicting 3D structure in terms of PBs with a Bayesian approach yielded an accuracy rate of 36.0% for all loops and 41.2% for the short loops. Specific learning in the short loops increased the latter by 1%. CONCLUSION: This work highlights the difficulties of assigning repetitive structures and the advantages of using more precise descriptions, that is, PBs. We observed some new amino acid distributions in the short loops and used this information to enhance local prediction. Instead of describing entire loops, our approach predicts each position in the loops locally. It can thus be used to propose many different structures for the loops and to probe and sample their flexibility. It can be a useful tool in ab initio loop prediction.

Computational Biology↗
PubMed2000

Protein folding using contact maps.

We discuss the problem of representations of protein structure and give the definition of contact maps. We present a method to obtain a three-dimensional polypeptide conformation from a contact map. We also explain how to deal with the case of nonphysical contact maps. We describe a stochastic method to perform dynamics in contact map space. We explain how the motion is restricted to physical regions of the space. First, we introduce the exact free energy of a contact map and discuss two simple approximations to it. Second, we present a method to derive energy parameters based on perception learning. We prove in an extensive number of situations that the pairwise contact approximation both when alone and when supplemented with a hydrophobic term is unsuitable for stabilizing proteins' native states.

Aprotinin↗
PubMed2026

HyLnc: a hybrid deep learning and feature-based approach for long non-coding RNA prediction.

Long non-coding RNAs (lncRNAs) play important roles in gene regulation, development and disease, yet accurate identification of lncRNAs from transcriptomic data remains a major computational challenge. Existing methods often rely either on handcrafted sequence features or deep learning approaches, each with their inherent limitations in capturing the full complexity of RNA sequences. In this study, we proposed HyLnc, a computational framework that integrates transformer-based contextual embeddings with biologically meaningful sequence features for improved lncRNA prediction. A custom BERT-based model was first pre-trained on a large corpus of metazoan RNA sequences using a masked language modelling strategy to learn contextual nucleotide dependencies. The model was subsequently fine-tuned on curated datasets of lncRNAs and protein-coding transcripts and 256-dimensional deep sequence embeddings were extracted. Parallelly, 348 handcrafted features, including ORF characteristics, untranslated region (UTR) properties, nucleotide composition and Fickett scores, were computed. A multi-stage feature selection strategy was applied to identify the most informative features, resulting in optimized hybrid feature sets. Multiple machine learning classifiers were evaluated, with the RF model achieving the best performance. The proposed framework attained an accuracy of 91.30%, F1-score of 91.23% and MCC of 82.60 on an independent validation dataset, outperforming several existing lncRNA prediction tools. Thus, HyLnc demonstrates that integrating deep contextual representations with biologically interpretable features enhances lncRNA prediction. This approach provides a robust and scalable solution for large-scale transcriptome annotation and can be extended to other sequence-based prediction.

RNA, Long Noncoding↗
PubMed2000

Toward an energy function for the contact map representation of proteins.

We analyzed several energy functions for predicting the native state of proteins from an energy minimization procedure. We derived the parameters of a given energy function by imposing the basic requirement that the energy of the native conformation of a protein is lower than that of any conformation chosen from a set of decoys. Our work is motivated by a recent result which proved that the simple pairwise contact approximation of the energy is insufficient to satisfy simultaneously such a basic requirement for all the proteins in a database. Here, we investigate the reasons of such negative results and show how to improve the predictive power of methods based on energy minimization. We generated decoys by gapless threading, and we derive energy parameters by perceptron learning. We first considered hydrophobic contributions to the energy, defined in several ways, and showed that the additional hydrophobic terms enlarge slightly the number of proteins that can be stabilized together. Next, we performed various modifications of the pairwise energy term. We introduced (1) a distinction between inter-residue contacts on the surface and in the core of a protein and (2) a simple distance-dependent pairwise interaction in which a two-tier definition of contact replaces the original (single-tier) one. Our results suggest that a detailed treatment of the pairwise potential is likely to be more relevant than the consideration of other forces.

Algorithms↗
PubMed1998

Abnormal hippocampal spatial representations in alphaCaMKIIT286A and CREBalphaDelta- mice.

Hippocampal "place cells" fire selectively when an animal is in a specific location. The fine-tuning and stability of place cell firing was compared in two types of mutant mice with different long-term potentiation (LTP) and place learning impairments. Place cells from both mutants showed decreased spatial selectivity. Place cell stability was also deficient in both mutants and, consistent with the severities in their LTP and spatial learning deficits, was more affected in mice with a point mutation [threonine (T) at position 286 mutated to alanine (A)] in the alpha calmodulin kinase II (alphaCaMKIIT286A) than in mice deficient for the alpha and Delta isoforms of adenosine 3'5'-monophosphate-responsive element binding proteins (CREBalphaDelta-). Thus, LTP appears to be important for the fine tuning and stabilization of place cells, and these place cell properties may be necessary for spatial learning.

Animals↗
PubMed2004

A statistical framework for genomic data fusion.

MOTIVATION: During the past decade, the new focus on genomics has highlighted a particular challenge: to integrate the different views of the genome that are provided by various types of experimental data. RESULTS: This paper describes a computational framework for integrating and drawing inferences from a collection of genome-wide measurements. Each dataset is represented via a kernel function, which defines generalized similarity relationships between pairs of entities, such as genes or proteins. The kernel representation is both flexible and efficient, and can be applied to many different types of data. Furthermore, kernel functions derived from different types of data can be combined in a straightforward fashion. Recent advances in the theory of kernel methods have provided efficient algorithms to perform such combinations in a way that minimizes a statistical loss function. These methods exploit semidefinite programming techniques to reduce the problem of finding optimizing kernel combinations to a convex optimization problem. Computational experiments performed using yeast genome-wide datasets, including amino acid sequences, hydropathy profiles, gene expression data and known protein-protein interactions, demonstrate the utility of this approach. A statistical learning algorithm trained from all of these data to recognize particular classes of proteins--membrane proteins and ribosomal proteins--performs significantly better than the same algorithm trained on any single type of data. AVAILABILITY: Supplementary data at http://noble.gs.washington.edu/proj/sdp-svm

Algorithms↗
PubMed2005

A support vector machine approach to the identification of phosphorylation sites.

We describe a bioinformatics tool that can be used to predict the position of phosphorylation sites in proteins based only on sequence information. The method uses the support vector machine (SVM) statistical learning theory. The statistical models for phosphorylation by various types of kinases are built using a dataset of short (9-amino acid long) sequence fragments. The sequence segments are dissected around post-translationally modified sites of proteins that are on the current release of the Swiss-Prot database, and that were experimentally confirmed to be phosphorylated by any kinase. We represent them as vectors in a multidimensional abstract space of short sequence fragments. The prediction method is as follows. First, a given query protein sequence is dissected into overlapping short segments. All the fragments are then projected into the multidimensional space of sequence fragments via a collection of different representations. Those points are classified with pre-built statistical models (the SVM method with linear, polynomial and radial kernel functions) either as phosphorylated or inactive ones. The resulting list of plausible sites for phosphorylation by various types of kinases in the query protein is returned to the user. The efficiency of the method for each type of phosphorylation is estimated using leave-one-out tests and presented here. The sensitivities of the models can reach over 70%, depending on the type of kinase. The additional information from profile representations of short sequence fragments helps in gaining a higher degree of accuracy in some phosphorylation types. The further development of an automatic phosphorylation site annotation predictor based on our algorithm should yield a significant improvement when using statistical algorithms in order to quantify the results.

Algorithms↗
PubMed2026

Artificial Intelligence for Natural Products Discovery and Development.

Natural products (NPs) remain a cornerstone of modern drug discovery, offering stereochemical complexity and diverse bioactivities that precisely modulate therapeutic targets, refined through billions of years of evolution. However, their research has long been hindered by inefficient, empirical workflows, high resource consumption, structural complexity, and the "multicomponent, multi-target" nature of their mechanisms. The exponential growth of genomic, metabolomic, and spectral data has overwhelmed conventional analytical methods, exposing critical bottlenecks in handling high-dimensional, heterogeneous datasets that exceed human interpretive capacity. Artificial intelligence (AI) is emerging as a transformative paradigm to address these challenges, integrating multi-omics and chemical data to shift NP research from fragmented empiricism toward mechanism-driven, precision-oriented development. By leveraging deep learning architectures- including graph neural networks, Transformers, and diffusion-based generative models-AI enables systematic decoding of NP biosynthesis, automated structure elucidation, rational target identification, knowledge extraction from vast unstructured scientific literature, and de novo molecular design. This review comprehensively surveys recent advances in AI applications across the full NP discovery and development pipeline, encompassing genome mining, structure-based and ligand-based virtual screening, multimodal structural characterization, lead optimization, and biosynthetic pathway engineering. We further examine the emerging roles of protein-centric, molecule- centric, and multimodal foundation models, as well as large language models, in bridging genotype-to-chemotype gaps and unlocking unstructured scientific knowledge. Finally, we discuss critical challenges including data scarcity, representational limitations for complex stereochemistry, physical plausibility in generative models, and the urgent need for experimental validation, while outlining future directions toward autonomous experimentation, closed-loop optimization, and human-AI collaborative discovery.

Artificial intelligence↗
PubMed2004

An analysis of the neural representation of birdsong memory.

Songbirds, such as zebra finches, learn their song from a tutor early in life. Forebrain nuclei in the "song system" are important for the acquisition and production of song. Brain regions [including the caudomedial part of the neostriatum (NCM) and of the hyperstriatum ventrale (CMHV)] outside the song system show increased neuronal activation, measured as expression of immediate early genes (IEGs), when zebra finch males are exposed to song. IEG expression in the NCM in response to tutor song is significantly positively correlated with the strength of song learning (i.e., the number of elements copied). Here, we exposed three groups of adult zebra finch males to tutor song, to their own song, or to novel conspecific song. The two control groups were included to examine an alternative explanation of our previous results in terms of variation in predisposed levels of attentiveness. Expression of Zenk, the protein product of the IEG ZENK, was measured in the NCM, CMHV, and hippocampus. There were no significant differences in overall Zenk expression between the three experimental groups. However, there was a significant positive correlation between Zenk expression in the NCM (but not in the other two regions) and strength of song learning in the males that were exposed to the tutor song. There was no such correlation in the other two groups. These results suggest that experience-related neuronal activation is specific to the tutor song and thus unlikely to be a result of differences in attention.

Analysis of Variance↗
PubMed1994

Predicting location and structure of beta-sheet regions using stochastic tree grammars.

We describe and demonstrate the effectiveness of a method of predicting protein secondary structures, beta-sheet regions in particular, using a class of stochastic tree grammars as representational language for their amino acid sequence patterns. The family of stochastic tree grammars we use, the Stochastic Ranked Node Rewriting Grammars (SRNRG), is one of the rare families of stochastic grammars that are expressive enough to capture the kind of long-distance dependencies exhibited by the sequences of beta-sheet regions, and at the same time enjoy relatively efficient processing. We applied our method on real data obtained from the HSSP database and the results obtained are encouraging: Using an SRNRG trained by data of a particular protein, our method was actually able to predict the location and structure of beta-sheet regions in a number of different proteins, whose sequences are less than 25 per cent homologous to the training sequences. The learning algorithm we use is an extension of the 'Inside-Outside' algorithm for stochastic context free grammars, but with a number of significant modifications. First, we restricted the grammars used to be members of the 'linear' subclass of SRNRG, and devised simpler and faster algorithms for this subclass. Secondly, we reduced the alphabet size (i.e. the number of amino acids) by clustering them using their physicochemical properties, gradually through the iterations of the learning algorithm. Finally, we parallelized our parsing algorithm to run on a highly parallel computer, a 32-processor CM-5, and were able to obtain a nearly linear speed-up.(ABSTRACT TRUNCATED AT 250 WORDS)

Algorithms↗
PubMed2004

The long-term stability of new hippocampal place fields requires new protein synthesis.

The hippocampus is critical for formation of spatial memories. Hippocampal pyramidal neurons in freely behaving animals exhibit spatially selective firing patterns, which taken together form an internal representation of the environment. This representation is thought to contribute to the hippocampal spatial memory system. Behavioral long-term memories differ from short-term memories in requiring the synthesis of new proteins. Does the development of the internal hippocampal representation also require the synthesis of new proteins? We found that blocking protein synthesis in the brain of mice by 95% does not affect short-term stability of newly formed hippocampal place fields but abolishes stability in the long term. By contrast, inhibiting protein synthesis does not affect the retention and recall of previously established fields in a familiar environment, indicating that protein synthesis-dependent reconsolidation is not required for recall. Our results indicate that place fields parallel both behavioral memories and the late phase of long-term potentiation in requiring the synthesis of new proteins for consolidation.

Action Potentials↗
PubMed2004

Altered representation of the spatial code for odors after olfactory classical conditioning; memory trace formation by synaptic recruitment.

In the olfactory bulb of vertebrates or the homologous antennal lobe of insects, odor quality is represented by stereotyped patterns of neuronal activity that are reproducible within and between individuals. Using optical imaging to monitor synaptic activity in the Drosophila antennal lobe, we show here that classical conditioning rapidly alters the neural code representing the learned odor by recruiting new synapses into that code. Pairing of an odor-conditioned stimulus with an electric shock-unconditioned stimulus causes new projection neuron synapses to respond to the odor along with those normally activated prior to conditioning. Different odors recruit different groups of projection neurons into the spatial code. The change in odor representation after conditioning appears to be intrinsic to projection neurons. The rapid recruitment by conditioning of new synapses into the representation of sensory information may be a general mechanism underlying many forms of short-term memory.

Animals↗