Automated assembly of molecular mechanisms at scale from text mining and curated databases

ENQUIRE automatically reconstructs, expands, and drives enrichment analysis of gene and Mesh co-occurrence networks from context-specific biomedical literature

The accelerating growth of scientific literature overwhelms our capacity to manually distil complex phenomena like molecular networks linked to diseases. Moreover, biases in biomedical research and database annotation limit our interpretation of facts and generation of hypotheses. ENQUIRE (Expanding Networks by Querying Unexpectedly Inter-Related Entities) offers a time- and resource-efficient alternative to manual literature curation and database mining. ENQUIRE reconstructs and expands co-occurrence networks of genes and biomedical ontologies from user-selected input corpora and network-inferred PubMed queries. Its modest resource usage and the integration of text mining, automatic querying, and network-based statistics mitigating literature biases makes ENQUIRE unique in its broad-scope applications. For example, ENQUIRE can generate co-occurrence gene networks that reflect high-confidence, functional networks. When tested on case studies spanning cancer, cell differentiation, and immunity, ENQUIRE identified interlinked genes and enriched pathways unique to each topic, thereby preserving their underlying context specificity. ENQUIRE supports biomedical researchers by easing literature annotation, boosting hypothesis formulation, and facilitating the identification of molecular targets for subsequent experimentation.

A Computational Protocol for the Knowledge-Based Assessment and Capture of Pathologies

We propose that one of the main hurdles in delivering comprehensively informed care results from the challenges surrounding the extraction, representation, and retention of prior clinical experience and basic medical knowledge, as well as its translation into time- and context-informed actionable interventions. While emerging applications in artificial intelligence-based techniques, for example, large language models, offer impressive pattern association capabilities, they often fall short in producing human-readable explanations crucial to their integration into clinical care. Moreover, they require large well-defined and well-integrated data sets that typically conflict with the availability of such data in all but a few areas of medicine, for example, medical imaging and neuroimaging, noninvasive monitoring of bio-electrical activity, etc. In this chapter, we argue that approximate reasoning rooted in the knowledge that is explainable to the human clinician may offer attractive avenues for the introduction of such knowledge in a systematic way that supports formal retention, sharing, and reuse of new clinical and basic medical experience. We outline a conceptual protocol that targets the use of sparse and disparate data of different types and from different sources, seamlessly drawing on our collective experience and that of others. We illustrate the utility of such an integrative approach by applying the latter to the assessment and reconciliation of data from different experimental models, human and animal, in the example use case of a complex health condition.

Protein-Protein Interaction Networks Derived from Classical and Machine Learning-Based Natural Language Processing Tools

The study of protein-protein interactions (PPIs) provides insight into various biological mechanisms, including the binding of antibodies to antigens, enzymes to inhibitors or promoters, and receptors to ligands. Recent studies of PPIs have led to significant biological breakthroughs. For example, the study of PPIs involved in the human:SARS-CoV-2 viral infection mechanism aided in the development of SARS-CoV-2 vaccines. Though several databases exist for the manual curation of PPI networks, text mining methods have been routinely demonstrated as useful alternatives for newly studied or understudied species, where databases are incomplete. Here, the relationship extraction performance of several open-source classical text processing, machine learning (ML)-based natural language processing (NLP), and large language model (LLM)-based NLP tools was compared. Overall, our results indicated that networks derived from classical methods tend to have high true positive rates at the expense of having overconnected networks, ML-based NLP methods have lower true positive rates but networks with the closest structures to the target network, and LLM-based NLP methods tend to exist between the two other approaches, with variable performances. The selection of a specific NLP approach should be tied to the needs of a study and text availability, as models varied in performance due to the amount of text provided.

Eliater: a Python package for estimating outcomes of perturbations in biomolecular networks

SUMMARY

We introduce Eliater, a Python package for estimating the effect of perturbation of an upstream molecule on a downstream molecule in a biomolecular network. The estimation takes as input a biomolecular network, observational biomolecular data, and a perturbation of interest, and outputs an estimated quantitative effect of the perturbation. We showcase the functionalities of Eliater in a case study of Escherichia coli transcriptional regulatory network.

AVAILABILITY AND IMPLEMENTATION

The code, the documentation, and several case studies are available open source at https://github.com/y0-causal-inference/eliater.

Nociceptor-immune interactomes reveal insult-specific immune signatures of pain

Inflammatory pain results from the heightened sensitivity and reduced threshold of nociceptor sensory neurons due to exposure to inflammatory mediators. However, the cellular and transcriptional diversity of immune cell and sensory neuron types makes it challenging to decipher the immune mechanisms underlying pain. Here we used single-cell transcriptomics to determine the immune gene signatures associated with pain development in three skin inflammatory pain models in mice: zymosan injection, skin incision and ultraviolet burn. We found that macrophage and neutrophil recruitment closely mirrored the kinetics of pain development and identified cell-type-specific transcriptional programs associated with pain and its resolution. Using a comprehensive list of potential interactions mediated by receptors, ligands, ion channels and metabolites to generate injury-specific neuroimmune interactomes, we also uncovered that thrombospondin-1 upregulated by immune cells upon injury inhibited nociceptor sensitization. This study lays the groundwork for identifying the neuroimmune axes that modulate pain in diverse disease contexts.

drGAT: Attention-Guided Gene Assessment of Drug Response Utilizing a Drug-Cell-Gene Heterogeneous Network

Drug development is a lengthy process with a high failure rate. Increasingly, machine learning is utilized to facilitate the drug development processes. These models aim to enhance our understanding of drug characteristics, including their activity in biological contexts. However, a major challenge in drug response (DR) prediction is model interpretability as it aids in the validation of findings. This is important in biomedicine, where models need to be understandable in comparison with established knowledge of drug interactions with proteins. drGAT, a graph deep learning model, leverages a heterogeneous graph composed of relationships between proteins, cell lines, and drugs. drGAT is designed with two objectives: DR prediction as a binary sensitivity prediction and elucidation of drug mechanism from attention coefficients. drGAT has demonstrated superior performance over existing models, achieving 78% accuracy (and precision), and 76% F1 score for 269 DNA-damaging compounds of the NCI60 drug response dataset. To assess the model's interpretability, we conducted a review of drug-gene co-occurrences in Pubmed abstracts in comparison to the top 5 genes with the highest attention coefficients for each drug. We also examined whether known relationships were retained in the model by inspecting the neighborhoods of topoisomerase-related drugs. For example, our model retained TOP1 as a highly weighted predictive feature for irinotecan and topotecan, in addition to other genes that could potentially be regulators of the drugs. Our method can be used to accurately predict sensitivity to drugs and may be useful in the identification of biomarkers relating to the treatment of cancer patients.

Extending BioMASS to construct mathematical models from external knowledge

Motivation

Mechanistic modeling based on ordinary differential equations has led to numerous findings in systems biology by integrating prior knowledge and experimental data. However, the manual curation of knowledge necessary when constructing models poses a bottleneck. As the speed of knowledge accumulation continues to grow, there is a demand for a scalable means of constructing executable models.

Results

We previously introduced BioMASS-an open-source, Python-based framework-to construct, simulate, and analyze mechanistic models of signaling networks. With one of its features, Text2Model, BioMASS allows users to define models in a natural language-like format, thereby facilitating the construction of large-scale models. We demonstrate that Text2Model can serve as a tool for integrating external knowledge for mathematical modeling by generating Text2Model files from a pathway database or through the use of a large language model, and simulating its dynamics through BioMASS. Our findings reveal the tool's capabilities to encourage exploration from prior knowledge and pave the way for a fully data-driven approach to constructing mathematical models.

Availability and implementation

The code and documentation for BioMASS are available at https://github.com/biomass-dev/biomass and https://biomass-core.readthedocs.io, respectively. The code used in this article are available at https://github.com/okadalabipr/text2model-from-knowledge.

BioKC: a collaborative platform for curation and annotation of molecular interactions

Curation of biomedical knowledge into systems biology diagrammatic or computational models is essential for studying complex biological processes. However, systems-level curation is a laborious manual process, especially when facing ever-increasing growth of domain literature. New findings demonstrating elaborate relationships between multiple molecules, pathways and cells have to be represented in a format suitable for systems biology applications. Importantly, curation should capture the complexity of molecular interactions in such a format together with annotations of the involved elements and support stable identifiers and versioning. This challenge calls for novel collaborative tools and platforms allowing to improve the quality and the output of the curation process. In particular, community-based curation, an important source of curated knowledge, requires support in role management, reviewing features and versioning. Here, we present Biological Knowledge Curation (BioKC), a web-based collaborative platform for the curation and annotation of biomedical knowledge following the standard data model from Systems Biology Markup Language (SBML). BioKC offers a graphical user interface for curation of complex molecular interactions and their annotation with stable identifiers and supporting sentences. With the support of collaborative curation and review, it allows to construct building blocks for systems biology diagrams and computational models. These building blocks can be published under stable identifiers and versioned and used as annotations, supporting knowledge building for modelling activities.

Illuminating function of the understudied druggable kinome

The human kinome, with more than 500 proteins, is crucial for cell signaling and disease. Yet, about one-third of kinases lack in-depth study. The Data and Resource Generating Center for Understudied Kinases has developed multiple resources to address this challenge including creation of a heavy amino acid peptide library for parallel reaction monitoring and quantitation of protein kinase expression, use of understudied kinases tagged with a miniTurbo-biotin ligase to determine interaction networks by proximity-dependent protein biotinylation, NanoBRET probe development for screening chemical tool target specificity in live cells, characterization of small molecule chemical tools inhibiting understudied kinases, and computational tools for defining kinome architecture. These resources are available through the Dark Kinase Knowledgebase, supporting further research into these understudied protein kinases.

MatKG: An autonomously generated knowledge graph in Material Science

In this paper, we present MatKG, a knowledge graph in materials science that offers a repository of entities and relationships extracted from scientific literature. Using advanced natural language processing techniques, MatKG includes an array of entities, including materials, properties, applications, characterization and synthesis methods, descriptors, and symmetry phase labels. The graph is formulated based on statistical metrics, encompassing over 70,000 entities and 5.4 million unique triples. To enhance accessibility and utility, we have serialized MatKG in both CSV and RDF formats and made these, along with the code base, available to the research community. As the largest knowledge graph in materials science to date, MatKG provides structured organization of domain-specific data. Its deployment holds promise for various applications, including material discovery, recommendation systems, and advanced analytics.

reguloGPT: Harnessing GPT for Knowledge Graph Construction of Molecular Regulatory Pathways

Motivation

Molecular Regulatory Pathways (MRPs) are crucial for understanding biological functions. Knowledge Graphs (KGs) have become vital in organizing and analyzing MRPs, providing structured representations of complex biological interactions. Current tools for mining KGs from biomedical literature are inadequate in capturing complex, hierarchical relationships and contextual information about MRPs. Large Language Models (LLMs) like GPT-4 offer a promising solution, with advanced capabilities to decipher the intricate nuances of language. However, their potential for end-to-end KG construction, particularly for MRPs, remains largely unexplored.

Results

We present reguloGPT, a novel GPT-4 based in-context learning prompt, designed for the end-to-end joint name entity recognition, N-ary relationship extraction, and context predictions from a sentence that describes regulatory interactions with MRPs. Our reguloGPT approach introduces a context-aware relational graph that effectively embodies the hierarchical structure of MRPs and resolves semantic inconsistencies by embedding context directly within relational edges. We created a benchmark dataset including 400 annotated PubMed titles on N6-methyladenosine (m6A) regulations. Rigorous evaluation of reguloGPT on the benchmark dataset demonstrated marked improvement over existing algorithms. We further developed a novel G-Eval scheme, leveraging GPT-4 for annotation-free performance evaluation and demonstrated its agreement with traditional annotation-based evaluations. Utilizing reguloGPT predictions on m6A-related titles, we constructed the m6A-KG and demonstrated its utility in elucidating m6A's regulatory mechanisms in cancer phenotypes across various cancers. These results underscore reguloGPT's transformative potential for extracting biological knowledge from the literature.

Availability and implementation

The source code of reguloGPT, the m6A title and benchmark datasets, and m6A-KG are available at: https://github.com/Huang-AI4Medicine-Lab/reguloGPT.

Knowledge Graphs and Their Applications in Drug Discovery

Knowledge graphs represent information in the form of entities and relationships between those entities. Such a representation has multiple potential applications in drug discovery, including democratizing access to biomedical data, contextualizing or visualizing that data, and generating novel insights through the application of machine learning approaches. Knowledge graphs put data into context and therefore offer the opportunity to generate explainable predictions, which is a key topic in contemporary artificial intelligence. In this chapter, we outline some of the factors that need to be considered when constructing biomedical knowledge graphs, examine recent advances in mining such systems to gain insights for drug discovery, and identify potential future areas for further development.

Multi-scale models of whole cells: progress and challenges

Whole-cell modeling is "the ultimate goal" of computational systems biology and "a grand challenge for 21st century" (Tomita, Trends in Biotechnology, 2001, 19(6), 205-10). These complex, highly detailed models account for the activity of every molecule in a cell and serve as comprehensive knowledgebases for the modeled system. Their scope and utility far surpass those of other systems models. In fact, whole-cell models (WCMs) are an amalgam of several types of "system" models. The models are simulated using a hybrid modeling method where the appropriate mathematical methods for each biological process are used to simulate their behavior. Given the complexity of the models, the process of developing and curating these models is labor-intensive and to date only a handful of these models have been developed. While whole-cell models provide valuable and novel biological insights, and to date have identified some novel biological phenomena, their most important contribution has been to highlight the discrepancy between available data and observations that are used for the parametrization and validation of complex biological models. Another realization has been that current whole-cell modeling simulators are slow and to run models that mimic more complex (e.g., multi-cellular) biosystems, those need to be executed in an accelerated fashion on high-performance computing platforms. In this manuscript, we review the progress of whole-cell modeling to date and discuss some of the ways that they can be improved.

Bacterial microbiota in different types of processed meat products: diversity, adaptation, and co-occurrence

As a double-edged sword, some bacterial microbes can improve the quality and shelf life of meat products, but others mainly responsible for deterioration of the safety and quality of meat products. This review aims to present a landscape of the bacterial microbiota in different types of processed meat products. After demonstrating a panoramic view of the bacterial genera in meat products, the diversity of bacterial microbiota was evaluated in two dimensions, namely different types of processed meat products and different meats. Then, the influence of environmental factors on bacterial communities was evaluated according to the storage temperature, packaging conditions, and sterilization methods. Furthermore, microbes are not independent. To explore interactions among those genera, co-occurrence patterns were examined. In these respects, this review highlighted the recent advances in fundamental principles that underlie the environmental adaption tricks and why some species tend to occur together frequently, such as metabolic cross-feeding, co-aggregate at microscale, and the intercellular signaling system. Further investigations are required to unveil the underlying molecular mechanisms that govern microbial community systems, ultimately contributing to developing new strategies to harness beneficial microorganisms and control harmful microorganisms.

A regulatory pathway model of neuropsychological disruption in Havana syndrome

Introduction

In 2016 diplomatic personnel serving in Havana, Cuba, began reporting audible sensory phenomena paired with onset of complex and persistent neurological symptoms consistent with brain injury. The etiology of these Anomalous Health Incidents (AHI) and subsequent symptoms remains unknown. This report investigates putative exposure-symptom pathology by assembling a network model of published bio-behavioral pathways and assessing how dysregulation of such pathways might explain loss of function in these subjects using data available in the published literature. Given similarities in presentation with mild traumatic brain injury (mTBI), we used the latter as a clinically relevant means of evaluating if the neuropsychological profiles observed in Havana Syndrome Havana Syndrome might be explained at least in part by a dysregulation of neurotransmission, neuro-inflammation, or both.

Method

Automated text-mining of >9,000 publications produced a network consisting of 273 documented regulatory interactions linking 29 neuro-chemical markers with 9 neuropsychological constructs from the Brief Mood Survey, PTSD Checklist, and the Frontal Systems Behavior Scale. Analysis of information flow through this network produced a set of regulatory rules reconciling to within a 6% departure known mechanistic pathways with neuropsychological profiles in N = 6 subjects.

Results

Predicted expression of neuro-chemical markers that jointly satisfy documented pathways and observed symptom profiles display characteristically elevated IL-1B, IL-10, NGF, and norepinephrine levels in the context of depressed BDNF, GDNF, IGF1, and glutamate expression (FDR < 5%). Elevations in CRH and IL-6 were also predicted unanimously across all subjects. Furthermore, simulations of neurological regulatory dynamics reveal subjects do not appear to be "locked in" persistent illness but rather appear to be engaged in a slow recovery trajectory.

Discussion

This computational analysis of measured neuropsychological symptoms in Havana-based diplomats proposes that these AHI symptoms may be supported in part by disruption of known neuroimmune and neurotransmission regulatory mechanisms also associated with mTBI.