MOTIVATION: Large language models (LLMs) are being adopted at an unprecedented rate, yet still face challenges in knowledge-intensive domains such as biomedicine. Solutions such as pretraining and domain-specific fine-tuning add substantial computational overhead, requiring further domain-expertise. Here, we introduce a token-optimized and robust Knowledge Graph-based Retrieval Augmented Generation (KG-RAG) framework by leveraging a massive biomedical KG (SPOKE) with LLMs such as Llama-2-13b, GPT-3.5-Turbo, and GPT-4, to generate meaningful biomedical text rooted in established knowledge. RESULTS: Compared to the existing RAG technique for Knowledge Graphs, the proposed method utilizes minimal graph schema for context extraction and uses embedding methods for context pruning. This optimization in context extraction results in more than 50% reduction in token consumption without compromising the accuracy, making a cost-effective and robust RAG implementation on proprietary LLMs. KG-RAG consistently enhanced the performance of LLMs across diverse biomedical prompts by generating responses rooted in established knowledge, accompanied by accurate provenance and statistical evidence (if available) to substantiate the claims. Further benchmarking on human curated datasets, such as biomedical true/false and multiple-choice questions (MCQ), showed a remarkable 71% boost in the performance of the Llama-2 model on the challenging MCQ dataset, demonstrating the framework's capacity to empower open-source models with fewer parameters for domain-specific questions. Furthermore, KG-RAG enhanced the performance of proprietary GPT models, such as GPT-3.5 and GPT-4. In summary, the proposed framework combines explicit and implicit knowledge of KG and LLM in a token optimized fashion, thus enhancing the adaptability of general-purpose LLMs to tackle domain-specific questions in a cost-effective fashion. AVAILABILITY AND IMPLEMENTATION: SPOKE KG can be accessed at https://spoke.rbvi.ucsf.edu/neighborhood.html. It can also be accessed using REST-API (https://spoke.rbvi.ucsf.edu/swagger/). KG-RAG code is made available at https://github.com/BaranziniLab/KG_RAG. Biomedical benchmark datasets used in this study are made available to the research community in the same GitHub repository.

MOTIVATION: Knowledge graphs (KGs) are being adopted in industry, commerce and academia. Biomedical KG presents a challenge due to the complexity, size and heterogeneity of the underlying information. RESULTS: In this work, we present the Scalable Precision Medicine Open Knowledge Engine (SPOKE), a biomedical KG connecting millions of concepts via semantically meaningful relationships. SPOKE contains 27 million nodes of 21 different types and 53 million edges of 55 types downloaded from 41 databases. The graph is built on the framework of 11 ontologies that maintain its structure, enable mappings and facilitate navigation. SPOKE is built weekly by python scripts which download each resource, check for integrity and completeness, and then create a 'parent table' of nodes and edges. Graph queries are translated by a REST API and users can submit searches directly via an API or a graphical user interface. Conclusions/Significance: SPOKE enables the integration of seemingly disparate information to support precision medicine efforts. AVAILABILITY AND IMPLEMENTATION: The SPOKE neighborhood explorer is available at https://spoke.rbvi.ucsf.edu. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Since the initial publication of clusterMaker, the need for tools to analyze large biological datasets has only increased. New datasets are significantly larger than a decade ago, and new experimental techniques such as single-cell transcriptomics continue to drive the need for clustering or classification techniques to focus on portions of datasets of interest. While many libraries and packages exist that implement various algorithms, there remains the need for clustering packages that are easy to use, integrated with visualization of the results, and integrated with other commonly used tools for biological data analysis. clusterMaker2 has added several new algorithms, including two entirely new categories of analyses: node ranking and dimensionality reduction. Furthermore, many of the new algorithms have been implemented using the Cytoscape jobs API, which provides a mechanism for executing remote jobs from within Cytoscape. Together, these advances facilitate meaningful analyses of modern biological datasets despite their ever-increasing size and complexity. RESULTS: The use of clusterMaker2 is exemplified by reanalyzing the yeast heat shock expression experiment that was included in our original paper; however, here we explored this dataset in significantly more detail. Combining this dataset with the yeast protein-protein interaction network from STRING, we were able to perform a variety of analyses and visualizations from within clusterMaker2, including Leiden clustering to break the entire network into smaller clusters, hierarchical clustering to look at the overall expression dataset, dimensionality reduction using UMAP to find correlations between our hierarchical visualization and the UMAP plot, fuzzy clustering, and cluster ranking. Using these techniques, we were able to explore the highest-ranking cluster and determine that it represents a strong contender for proteins working together in response to heat shock. We found a series of clusters that, when re-explored as fuzzy clusters, provide a better presentation of mitochondrial processes. CONCLUSIONS: clusterMaker2 represents a significant advance over the previously published version, and most importantly, provides an easy-to-use tool to perform clustering and to visualize clusters within the Cytoscape network context. The new algorithms should be welcome to the large population of Cytoscape users, particularly the new dimensionality reduction and fuzzy clustering techniques.

SUMMARY: IntAct App is a Cytoscape 3 application that grants in-depth access to IntAct's molecular interaction data. It build networks where nodes are interacting molecules (mainly proteins, but also genes, RNA, chemicals…) and edges represent evidence of interaction. Users can query a network by providing its molecules, identified by different fields and optionally include all their interacting partners in the resulting network. The app offers three visualizations: one only displaying interactions, another representing every evidence and the last one emphasizing evidence where mutated versions of proteins were used. Users can also filter networks and click on nodes and edges to access all their related details. Finally, the application supports automation of its main features via Cytoscape commands. AVAILABILITY AND IMPLEMENTATION: Implementation available at https://apps.cytoscape.org/apps/intactapp, while the source code is available at https://github.com/EBI-IntAct/IntactApp.

Biological network figures are ubiquitous in the biology and medical literature. On the one hand, a good network figure can quickly provide information about the nature and degree of interactions between items and enable inferences about the reason for those interactions. On the other hand, good network figures are difficult to create. In this paper, we outline 10 simple rules for creating biological network figures for communication, from choosing layouts, to applying color or other channels to show attributes, to the use of layering and separation. These rules are accompanied by illustrative examples. We also provide a concise set of references and additional resources for each rule.

Peroxiredoxins (Prxs or Prdxs) are a large protein superfamily of antioxidant enzymes that rapidly detoxify damaging peroxides and/or affect signal transduction and, thus, have roles in proliferation, differentiation, and apoptosis. Prx superfamily members are widespread across phylogeny and multiple methods have been developed to classify them. Here we present an updated atlas of the Prx superfamily identified using a novel method called MISST (Multi-level Iterative Sequence Searching Technique). MISST is an iterative search process developed to be both agglomerative, to add sequences containing similar functional site features, and divisive, to split groups when functional site features suggest distinct functionally-relevant clusters. Superfamily members need not be identified initially-MISST begins with a minimal representative set of known structures and searches GenBank iteratively. Further, the method's novelty lies in the manner in which isofunctional groups are selected; rather than use a single or shifting threshold to identify clusters, the groups are deemed isofunctional when they pass a self-identification criterion, such that the group identifies itself and nothing else in a search of GenBank. The method was preliminarily validated on the Prxs, as the Prxs presented challenges of both agglomeration and division. For example, previous sequence analysis clustered the Prx functional families Prx1 and Prx6 into one group. Subsequent expert analysis clearly identified Prx6 as a distinct functionally relevant group. The MISST process distinguishes these two closely related, though functionally distinct, families. Through MISST search iterations, over 38,000 Prx sequences were identified, which the method divided into six isofunctional clusters, consistent with previous expert analysis. The results represent the most complete computational functional analysis of proteins comprising the Prx superfamily. The feasibility of this novel method is demonstrated by the Prx superfamily results, laying the foundation for potential functionally relevant clustering of the universe of protein sequences.

BACKGROUND: Development of automatable processes for clustering proteins into functionally relevant groups is a critical hurdle as an increasing number of sequences are deposited into databases. Experimental function determination is exceptionally time-consuming and can't keep pace with the identification of protein sequences. A tool, DASP (Deacon Active Site Profiler), was previously developed to identify protein sequences with active site similarity to a query set. Development of two iterative, automatable methods for clustering proteins into functionally relevant groups exposed algorithmic limitations to DASP. RESULTS: The accuracy and efficiency of DASP was significantly improved through six algorithmic enhancements implemented in two stages: DASP2 and DASP3. Validation demonstrated DASP3 provides greater score separation between true positives and false positives than earlier versions. In addition, DASP3 shows similar performance to previous versions in clustering protein structures into isofunctional groups (validated against manual curation), but DASP3 gathers and clusters protein sequences into isofunctional groups more efficiently than DASP and DASP2. CONCLUSIONS: DASP algorithmic enhancements resulted in improved efficiency and accuracy of identifying proteins that contain active site features similar to those of the query set. These enhancements provide incremental improvement in structure database searches and initial sequence database searches; however, the enhancements show significant improvement in iterative sequence searches, suggesting DASP3 is an appropriate tool for the iterative processes required for clustering proteins into isofunctional groups.

MOTIVATION: cddApp is a Cytoscape extension that supports the annotation of protein networks with information about domains and specific functional sites from the National Center for Biotechnology Information's conserved domain database (CDD). CDD information is loaded for nodes annotated with NCBI numbers or UniProt identifiers and (optionally) Protein Data Bank structures. cddApp integrates with the Cytoscape apps structureViz2 and enhancedGraphics. Together, these three apps provide powerful tools to annotate nodes with CDD domain and site information and visualize that information in both network and structural contexts. AVAILABILITY AND IMPLEMENTATION: cddApp is written in Java and freely available for download from the Cytoscape app store (http://apps.cytoscape.org). Documentation is provided at http://www.rbvi.ucsf.edu/cytoscape, and the source is publically available from GitHub http://github.com/RBVI/cddApp.

MOTIVATION: Clustering protein sequence data into functionally specific families is a difficult but important problem in biological research. One useful approach for tackling this problem involves representing the sequence dataset as a protein similarity network, and afterwards clustering the network using advanced graph analysis techniques. Although a multitude of such network clustering algorithms have been developed over the past few years, comparing algorithms is often difficult because performance is affected by the specifics of network construction. We investigate an important aspect of network construction used in analyzing protein superfamilies and present a heuristic approach for improving the performance of several algorithms. RESULTS: We analyzed how the performance of network clustering algorithms relates to thresholding the network prior to clustering. Our results, over four different datasets, show how for each input dataset there exists an optimal threshold range over which an algorithm generates its most accurate clustering output. Our results further show how the optimal threshold range correlates with the shape of the edge weight distribution for the input similarity network. We used this correlation to develop an automated threshold selection heuristic in order to most optimally filter a similarity network prior to clustering. This heuristic allows researchers to process their protein datasets with runtime efficient network clustering algorithms without sacrificing the clustering accuracy of the final results. AVAILABILITY: Python code for implementing the automated threshold selection heuristic, together with the datasets used in our analysis, are available at http://www.rbvi.ucsf.edu/Research/cytoscape/threshold_scripts.zip.

BACKGROUND: In the post-genomic era, the rapid increase in high-throughput data calls for computational tools capable of integrating data of diverse types and facilitating recognition of biologically meaningful patterns within them. For example, protein-protein interaction data sets have been clustered to identify stable complexes, but scientists lack easily accessible tools to facilitate combined analyses of multiple data sets from different types of experiments. Here we present clusterMaker, a Cytoscape plugin that implements several clustering algorithms and provides network, dendrogram, and heat map views of the results. The Cytoscape network is linked to all of the other views, so that a selection in one is immediately reflected in the others. clusterMaker is the first Cytoscape plugin to implement such a wide variety of clustering algorithms and visualizations, including the only implementations of hierarchical clustering, dendrogram plus heat map visualization (tree view), k-means, k-medoid, SCPS, AutoSOME, and native (Java) MCL. RESULTS: Results are presented in the form of three scenarios of use: analysis of protein expression data using a recently published mouse interactome and a mouse microarray data set of nearly one hundred diverse cell/tissue types; the identification of protein complexes in the yeast Saccharomyces cerevisiae; and the cluster analysis of the vicinal oxygen chelate (VOC) enzyme superfamily. For scenario one, we explore functionally enriched mouse interactomes specific to particular cellular phenotypes and apply fuzzy clustering. For scenario two, we explore the prefoldin complex in detail using both physical and genetic interaction clusters. For scenario three, we explore the possible annotation of a protein as a methylmalonyl-CoA epimerase within the VOC superfamily. Cytoscape session files for all three scenarios are provided in the Additional Files section. CONCLUSIONS: The Cytoscape plugin clusterMaker provides a number of clustering algorithms and visualizations that can be used independently or in combination for analysis and visualization of biological data sets, and for confirming or generating hypotheses about biological function. Several of these visualizations and algorithms are only available to Cytoscape users through the clusterMaker plugin. clusterMaker is available via the Cytoscape plugin manager.

SUMMARY: GLay provides Cytoscape users an assorted collection of versatile community structure algorithms and graph layout functions for network clustering and structured visualization. High performance is achieved by dynamically linking highly optimized C functions to the Cytoscape JAVA program, which makes GLay especially suitable for decomposition, display and exploratory analysis of large biological networks. AVAILABILITY: http://brainarray.mbni.med.umich.edu/glay/.

BACKGROUND: Leishmaniasis is a virulent parasitic infection that causes a worldwide disease burden. Most treatments have toxic side-effects and efficacy has decreased due to the emergence of resistant strains. The outlook is worsened by the absence of promising drug targets for this disease. We have taken a computational approach to the detection of new drug targets, which may become an effective strategy for the discovery of new drugs for this tropical disease. RESULTS: We have predicted the protein interaction network of Leishmania major by using three validated methods: PSIMAP, PEIMAP, and iPfam. Combining the results from these methods, we calculated a high confidence network (confidence score > 0.70) with 1,366 nodes and 33,861 interactions. We were able to predict the biological process for 263 interacting proteins by doing enrichment analysis of the clusters detected. Analyzing the topology of the network with metrics such as connectivity and betweenness centrality, we detected 142 potential drug targets after homology filtering with the human proteome. Further experiments can be done to validate these targets. CONCLUSION: We have constructed the first protein interaction network of the Leishmania major parasite by using a computational approach. The topological analysis of the protein network enabled us to identify a set of candidate proteins that may be both (1) essential for parasite survival and (2) without human orthologs. These potential targets are promising for further experimental validation. This strategy, if validated, may augment established drug discovery methodologies, for this and possibly other tropical diseases, with a relatively low additional investment of time and resources.

structureViz is a Cytoscape plug-in that links the visualization of biological networks provided by Cytoscape with the visualization and analysis of macromolecular structures and sequences provided by UCSF Chimera. When combined with Cytoscape and Chimera, structureViz provides the first tool that links these two critical aspects of computational analysis in a straightforward manner. structureViz includes commands to open structures in Chimera and align them using Chimera's sequence-structure analysis tools. When a structure is opened, structureViz provides an alternative interface to Chimera: the Cytoscape Molecular Structure Navigator. This interface uses a tree-based paradigm to allow users to select and affect the display of models, chains and residues, mostly through the use of context menus.