Integration Of Biological Data Research Articles

Predicting Response to [177Lu]Lu-PSMA Therapy in mCRPC Using Machine Learning

Background/Objectives: Radioligandtherapy (RLT) with [177Lu]Lu-PSMA has been newly introduced as a routine treatment for metastatic castration-resistant prostate cancer (mCRPC). However, not all patients can tolerate the entire therapeutic sequence, and in some cases, the treatment may prove ineffective. In real-world conditions, the aim is to distinguish between patients who fully benefit from treatment (those who respond effectively and tolerate the entire therapeutic sequence) and those who do not respond or cannot tolerate the entire sequence. This study explores predictive factors to distinguish between fully beneficial RLT treatment patients (FBTP) and not fully beneficial RLT treatment patients (NFBTP). The objective was to enhance the understanding of predictive factors influencing RLT effectiveness and to highlight the significance of machine learning in optimizing patient selection for treatment planning. Methods: Data from 25 mCRPC patients, categorized as FBTP (11) or NFBTP (14) to RLT, were analyzed. The dataset included clinical, imaging, and biological parameters. Data analysis techniques, including exploratory data analysis and feature engineering, were used to develop machine learning models for predicting patient outcomes. Results: Imaging data analysis revealed statistically significant differences in the renal uptake intensity of Choline between the two groups. A discordance of FDG+ and PSMA− was identified as a potential indicator of NFBTP. The integration of biological data enhanced the model’s predictive capability, achieving an accuracy of 0.92, a sensitivity of 0.96, and a precision of 0.96. Adding blood parameters like neutrophils, leukocytes, and alkaline phosphatase greatly increased prediction accuracy. Conclusions: This study emphasizes the significance of an integrated approach that merges imaging and biological data, thereby augmenting the predictive accuracy of patient outcomes in RLT with [177Lu]Lu-PSMA. In particular, including Choline PET among the imaging parameters provides unique insights into the predictive factors affecting RLT efficacy. This approach not only deepens the understanding of predictive factors but also underscores the utility of machine learning in refining the patient selection process for optimized treatment planning.

Detecting microsatellite instability by length comparison of microsatellites in the 3' untranslated region with RNA-seq.

Microsatellite instability (MSI), a phenomenon caused by deoxyribonucleic acid (DNA) mismatch repair system deficiencies, is an important biomarker in cancer research and clinical diagnostics. MSI detection often involves next-generation sequencing data, with many studies focusing on DNA. Here, we introduce a novel approach by measuring microsatellite lengths directly from ribonucleic acid sequencing (RNA-seq) data and comparing its distribution to detect MSI. Our findings reveal distinct instability patterns between MSI-high (MSI-H) and microsatellite stable samples, indicating the efficacy of RNA-based MSI detection. Additionally, microsatellites in the 3'-untranslated regions showed the greatest predictive value for MSI detection. Notably, this efficacy extends to detecting MSI-H samples even in tumors not commonly associated with MSI. Our approach highlights the utility of RNA-seq data in MSI detection, facilitating more precise diagnostics through the integration of various biological data.

TypeTaxonScript: sugarifying and enhancing data structures in biological systematics and biodiversity research.

Object-oriented programming (OOP) embodies a software development paradigm grounded in representing real-world entities as objects, facilitating a more efficient and structured modelling approach. In this article, we explore the synergy between OOP principles and the TypeScript (TS) programming language to create a JSON-formatted database designed for storing arrays of biological features. This fusion of technologies fosters a controlled and modular code script, streamlining the integration, manipulation, expansion, and analysis of biological data, all while enhancing syntax for improved human readability, such as through the use of dot notation. We advocate for biologists to embrace Git technology, akin to the practices of programmers and coders, for initiating versioned and collaborative projects. Leveraging the widely accessible and acclaimed IDE, Visual Studio Code, provides an additional advantage. Not only does it support running a Node.js environment, which is essential for running TS, but it also efficiently manages GitHub versioning. We provide a use case involving taxonomic data structure, focusing on angiosperm legume plants. This method is characterized by its simplicity, as the tools employed are both fully accessible and free of charge, and it is widely adopted by communities of professional programmers. Moreover, we are dedicated to facilitating practical implementation and comprehension through a comprehensive tutorial, a readily available pre-built database at GitHub, and a new package at npm.

Toward a Predictive Understanding of Cyanobacterial Harmful Algal Blooms through AI Integration of Physical, Chemical, and Biological Data.

Freshwater cyanobacterial harmful algal blooms (cyanoHABs) are a worldwide problem resulting in substantial economic losses, due to harm to drinking water supplies, commercial fishing, wildlife, property values, recreation, and tourism. Moreover, toxins produced from some cyanoHABs threaten human and animal health. Climate warming can affect the distribution of cyanoHABs, where rising temperatures facilitate more intense blooms and a greater distribution of cyanoHABs in inland freshwater. Nutrient runoff from adjacent watersheds is also a major driver of cyanoHAB formation. While some of the physicochemical factors behind cyanoHAB dynamics are known, there are still major gaps in our understanding of the conditions that trigger and sustain cyanoHABs over time. In this perspective, we suggest that sufficient data sets, as well as machine learning (ML) and artificial intelligence (AI) tools, are available to build a comprehensive model of cyanoHAB dynamics based on integrated environmental/climate, nutrient/water chemistry, and cyanoHAB microbiome and 'omics data to identify key factors contributing to HAB formation, intensity, and toxicity. By taking a holistic approach to the analysis of all available data, including the rapidly growing number of biological data sets, we can provide the foundational knowledge needed to address the increasing threat of cyanoHABs to the security of our water resources.

Use of Elasticsearch-based business intelligence tools for integration and visualization of biological data.

The emergence of massive datasets exploring the multiple levels of molecular biology has made their analysis and knowledge transfer more complex. Flexible tools to manage big biological datasets could be of great help for standardizing the usage of developed data visualizations and integration methods. Business intelligence (BI) tools have been used in many fields as exploratory tools. They have numerous connectors to link numerous data repositories with a unified graphic interface, offering an overview of data and facilitating interpretation for decision makers. BI tools could be a flexible and user-friendly way of handling molecular biological data with interactive visualizations. However, it is rather uncommon to see such tools used for the exploration of massive and complex datasets in biological fields. We believe that two main obstacles could be the reason. Firstly, we posit that the way to import data into BI tools are not compatible with biological databases. Secondly, BI tools may not be adapted to certain particularities of complex biological data, namely, the size, the variability of datasets and the availability of specialized visualizations. This paper highlights the use of five BI tools (Elastic Kibana, Siren Investigate, Microsoft Power BI, Salesforce Tableau and Apache Superset) onto which the massive data management repository engine called Elasticsearch is compatible. Four case studies will be discussed in which these BI tools were applied on biological datasets with different characteristics. We conclude that the performance of the tools depends on the complexity of the biological questions and the size of the datasets.

Omics data integration in computational biology viewed through the prism of machine learning paradigms.

Important quantities of biological data can today be acquired to characterize cell types and states, from various sources and using a wide diversity of methods, providing scientists with more and more information to answer challenging biological questions. Unfortunately, working with this amount of data comes at the price of ever-increasing data complexity. This is caused by the multiplication of data types and batch effects, which hinders the joint usage of all available data within common analyses. Data integration describes a set of tasks geared towards embedding several datasets of different origins or modalities into a joint representation that can then be used to carry out downstream analyses. In the last decade, dozens of methods have been proposed to tackle the different facets of the data integration problem, relying on various paradigms. This review introduces the most common data types encountered in computational biology and provides systematic definitions of the data integration problems. We then present how machine learning innovations were leveraged to build effective data integration algorithms, that are widely used today by computational biologists. We discuss the current state of data integration and important pitfalls to consider when working with data integration tools. We eventually detail a set of challenges the field will have to overcome in the coming years.

Deep-learning-enabled protein-protein interaction analysis for prediction of SARS-CoV-2 infectivity and variant evolution.

Host-pathogen interactions and pathogen evolution are underpinned by protein-protein interactions between viral and host proteins. An understanding of how viral variants affect protein-protein binding is important for predicting viral-host interactions, such as the emergence of new pathogenic SARS-CoV-2 variants. Here we propose an artificial intelligence-based framework called UniBind, in which proteins are represented as a graph at the residue and atom levels. UniBind integrates protein three-dimensional structure and binding affinity and is capable of multi-task learning for heterogeneous biological data integration. In systematic tests on benchmark datasets and further experimental validation, UniBind effectively and scalably predicted the effects of SARS-CoV-2 spike protein variants on their binding affinities to the human ACE2 receptor, as well as to SARS-CoV-2 neutralizing monoclonal antibodies. Furthermore, in a cross-species analysis, UniBind could be applied to predict host susceptibility to SARS-CoV-2 variants and to predict future viral variant evolutionary trends. This in silico approach has the potential to serve as an early warning system for problematic emerging SARS-CoV-2 variants, as well as to facilitate research on protein-protein interactions in general.

Abstract 5373: Autoencoder-based multimodal prediction of survival for non-small cell lung cancer

Abstract Most non-small cell lung cancer (NSCLC) prognosis prediction approaches use one data type and do not take advantage of the large amount of multimodal data available. To evaluate and explore the benefits of multimodal data integration, we present a combined feature selection and denoising autoencoder pipeline for NSCLC survival prediction and survival subtype identification using microRNA (miRNA), mRNA, DNA methylation, long non-coding RNA (lncRNA) and clinical data. Survival performance for both lung adenocarcinoma (LUAD) and squamous cell carcinoma (LUSC) patients was compared across modality combinations, data integration time and training data types. Multimodal data combinations outperformed single data modalities, with the early integration of all data modalities achieving concordance indexes (C-indexes) of 0.67 (±.04) and 0.63 (±.02) for LUAD and LUSC, respectively versus corresponding C-index of 0.64 (±.02) and 0.59 (±.03) for the best single cell modality (clinical). Notably, combining just lncRNA and clinical data facilitated effective survival discrimination, with C-indexes of 0.69 (±.03) for LUAD and 0.62 (±.03) for LUSC. Overall, higher performance was achieved by using a single denoising autoencoder for all biological data (early integration) and by training on both LUSC and LUAD patient data together. Two survival subtypes (log rank test p-value=1e-9) were identified, with 991 differentially expressed transcripts in the poorer survival group. Our analysis shows the value of multimodal data integration for predicting NSCLC progression, with especially good performance using the combination of lncRNA and clinical data. Early integration of biological data, with an initial linear feature selection technique and a denoising autoencoder for dimensionality reduction, showed effective survival performance and survival subtype identification. Further research is underway to expand analysis to different cancer types and data modalities and extract more biological interpretability from autoencoder models. Citation Format: Jacob G. Ellen, Etai Jacob, Nikos Nikolaou, Natasha Markuzon. Autoencoder-based multimodal prediction of survival for non-small cell lung cancer. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5373.

Insight into the Organization of the B10v3 Cucumber Genome by Integration of Biological and Bioinformatic Data

The availability of a well-organized and annotated reference genome is essential for genome research and the analysis of re-sequencing approaches. The B10v3 cucumber (Cucumis sativus L.) reference genome has been sequenced and assembled into 8035 contigs, a small fraction of which have been mapped to individual chromosomes. Currently, bioinformatics methods based on comparative homology have made it possible to re-order the sequenced contigs by mapping them to the reference genomes. The B10v3 genome (North-European, Borszczagowski line) was rearranged against the genomes of cucumber 9930 ('Chinese Long' line) and Gy14 (North American line). Furthermore, a better insight into the organization of the B10v3 genome was obtained by integrating the data available in the literature on the assignment of contigs to chromosomes in the B10v3 genome with the results of the bioinformatic analysis. The combination of information on the markers used in the assembly of the B10v3 genome and the results of FISH and DArT-seq experiments confirmed the reliability of the in silico assignment. Approximately 98% of the protein-coding genes within the chromosomes were assigned and a significant proportion of the repetitive fragments in the sequenced B10v3 genome were identified using the RagTag programme. In addition, BLAST analyses provided comparative information between the B10v3 genome and the 9930 and Gy14 data sets. This revealed both similarities and differences in the functional proteins found between the coding sequences region in the genomes. This study contributes to better knowledge and understanding of cucumber genome line B10v3.

CONTRABASS: exploiting flux constraints in genome-scale models for the detection of vulnerabilities.

Despite the fact that antimicrobial resistance is an increasing health concern, the pace of production of new drugs is slow due to the high cost and uncertain success of the process. The development of high-throughput technologies has allowed the integration of biological data into detailed genome-scale models of multiple organisms. Such models can be exploited by means of computational methods to identify system vulnerabilities such as chokepoint reactions and essential reactions. These vulnerabilities are appealing drug targets that can lead to novel drug developments. However, the current approach to compute these vulnerabilities is only based on topological data and ignores the dynamic information of the model. This can lead to misidentified drug targets. This work computes flux constraints that are consistent with a certain growth rate of the modelled organism, and integrates the computed flux constraints into the model to improve the detection of vulnerabilities. By exploiting these flux constraints, we are able to obtain a directionality of the reactions of metabolism consistent with a given growth rate of the model, and consequently, a more realistic detection of vulnerabilities can be performed. Several sets of reactions that are system vulnerabilities are defined and the relationships among them are studied. The approach for the detection of these vulnerabilities has been implemented in the Python tool CONTRABASS. Such tool, for which an online web server has also been implemented, computes flux constraints and generates a report with the detected vulnerabilities. CONTRABASS is available as an open source Python package at https://github.com/openCONTRABASS/CONTRABASS under GPL-3.0 License. An online web server is available at http://contrabass.unizar.es. A glossary of terms are available at Bioinformatics online.

A machine learning approach on whole blood immunomarkers to identify an inflammation-associated psychosis onset subgroup.

Psychosis onset is a transdiagnostic event that leads to a range of psychiatric disorders, which are currently diagnosed through clinical observation. The integration of multimodal biological data could reveal different subtypes of psychosis onset to target for the personalization of care. In this study, we tested the existence of subgroups of patients affected by first-episode psychosis (FEP) with a possible immunopathogenic basis. To do this, we designed a data-driven unsupervised machine learning model to cluster a sample of 127 FEP patients and 117 healthy controls (HC), based on the peripheral blood expression levels of 12 psychosis-related immune gene transcripts. To validate the model, we applied a resampling strategy based on the half-splitting of the total sample with random allocation of the cases. Further, we performed a post-hoc univariate analysis to verify the clinical, cognitive, and structural brain correlates of the subgroups identified. The model identified and validated two distinct clusters: 1) a FEP cluster characterized by the high expression of inflammatory and immune-activating genes (IL1B, CCR7, IL12A and CXCR3); 2) a cluster consisting of an equal number of FEP and HC subjects, which did not show a relative over or under expression of any immune marker (balanced subgroup). None of the subgroups was related to specific symptoms dimensions or longitudinal diagnosis of affective vs non-affective psychosis. FEP patients included in the balanced immune subgroup showed a thinning of the left supramarginal and superiorfrontal cortex (FDR-adjusted p-values < 0.05). Our results demonstrated the existence of a FEP patients' subgroup identified by a multivariate pattern of immunomarkers involved in inflammatory activation. This evidence may pave the way to sample stratification in clinical studies aiming to develop diagnostic tools and therapies targeting specific immunopathogenic pathways of psychosis.

From Biodiversity Observation Networks to Datasets and Workflows Supporting Biodiversity Indicators, a French Biodiversity Observation Network (BON) Essential Biodiversity Variables (EBV) Operationalization Pilot using Galaxy and Ecological Metadata Language

Integration of biological data with different ecological scales is complex! The biodiversity community (scientists, policy makers, managers, citizen, NGOs) needs to build a framework of harmonized and interoperable data from raw, heterogeneous and scattered datasets. Such a framework will help observation, measurement and understanding of the spatio-temporal dynamic of biodiversity from local to global scales. One of the most relevant approaches to reach that aim is the concept of Essential Biodiversity Variables (EBV). As we can potentially extract a lot of information from raw datasets sampled at different ecological scales, the EBV concept represents a useful leverage for identifying appropriate data to be collated as well as associated analytical workflows for processing these data. Thanks to FAIR data and source code implementation (Findable, Accessible, Interoperability, Reusable), it is possible to make a transparent assessment of biodiversity by generating operational biodiversity indicators (that can be reused / declined) through the EBV framework, and help designing or improving biodiversity monitoring at various scales. Through the BiodiFAIRse GO FAIR implementation network, we established how ecological and environmental sciences can benefit from existing open standards, tools and platforms used by European, Australian and United States infrastructures, particularly regarding the Galaxy platform for code sources accessiblility and the DataOne network of data catalogs and the Ecological Metadata Language standard for data management. We propose that these implementation choices can help fight the biodiversity crisis by supporting the important mission of GEO BON (Group on Earth Observation Biodiversity Observation Network): “Improve the acquisition, coordination and delivery of biodiversity observations and related services to users including decision makers and the scientific community” (GEO BON 2022).

CoVM2: Molecular Biological Data Integration of SARS-CoV-2 Proteins in a Macro-to-Micro Method.

The COVID-19 pandemic has been a major public health event since 2020. Multiple variant strains of SARS-CoV-2, the causative agent of COVID-19, were detected based on the mutation sites in their sequences. These sequence mutations may lead to changes in the protein structures and affect the binding states of SARS-CoV-2 and human proteins. Experimental research on SARS-CoV-2 has accumulated a large amount of structural data and protein-protein interactions (PPIs), but the studies on the SARS-CoV-2–human PPI networks lack integration of physical associations with possible protein docking information. In addition, the docking structures of variant viral proteins with human receptor proteins are still insufficient. This study constructed SARS-CoV-2–human protein–protein interaction network with data integration methods. Crystal structures were collected to map the interaction pairs. The pairs of direct interactions and physical associations were selected and analyzed for variant docking calculations. The study examined the structures of spike (S) glycoprotein of variants Delta B.1.617.2, Omicron BA.1, and Omicron BA.2. The calculated docking structures of S proteins and potential human receptors were obtained. The study integrated binary protein interactions with 3D docking structures to fulfill an extended view of SARS-CoV-2 proteins from a macro- to micro-scale.

Using real world data to advance the provision of supportive cancer care: mucositis as a case study.

For decades, clinical decision making and practice has been largely informed by data generated through randomized clinical trials (RCTs). By design, RCTs are highly restricted in both scope and scale, resulting in narrow indications and iterative advances in clinical practice. With the transition to electronic health records, there are now endless opportunities to utilize these 'real world' data (RWD) to make more substantive advances in our understanding that are, by nature, more applicable to reality. This review discusses the current paradigm of using big data to advance and inform the provision of supportive cancer care, using mucositis as a case study. Global efforts to synthesize RWD in cancer have almost exclusively focused on tumor classification and treatment efficacy, leveraging on routine tumor pathology and binary response outcomes. In contrast, clinical notes and billing codes are not as applicable to treatment side effects which require integration of both clinical and biological data, as well as patient-reported outcomes. Cancer treatment-induced toxicities are heterogeneous and complex, and as such, the use of RWD to better understand their etiology and interaction is challenging. Multidisciplinary cooperation and leadership are needed to improve data collection and governance to ensure the right data is accessible and reliable.

Integration of Heterogeneous Biological Data in Multiscale Mechanistic Model Calibration: Application to Lung Adenocarcinoma

Mechanistic models are built using knowledge as the primary information source, with well-established biological and physical laws determining the causal relationships within the model. Once the causal structure of the model is determined, parameters must be defined in order to accurately reproduce relevant data. Determining parameters and their values is particularly challenging in the case of models of pathophysiology, for which data for calibration is sparse. Multiple data sources might be required, and data may not be in a uniform or desirable format. We describe a calibration strategy to address the challenges of scarcity and heterogeneity of calibration data. Our strategy focuses on parameters whose initial values cannot be easily derived from the literature, and our goal is to determine the values of these parameters via calibration with constraints set by relevant data. When combined with a covariance matrix adaptation evolution strategy (CMA-ES), this step-by-step approach can be applied to a wide range of biological models. We describe a stepwise, integrative and iterative approach to multiscale mechanistic model calibration, and provide an example of calibrating a pathophysiological lung adenocarcinoma model. Using the approach described here we illustrate the successful calibration of a complex knowledge-based mechanistic model using only the limited heterogeneous datasets publicly available in the literature.

&lt;em&gt;In Silico&lt;/em&gt; Clinical Trials for Cardiovascular Disease

The SILICOFCM project mainly aims to develop a computational platform for in silico clinical trials of familial cardiomyopathies (FCMs). The unique characteristic of the platform is the integration of patient-specific biological, genetic, and clinical imaging data. The platform allows the testing and optimization of medical treatment to maximize positive therapeutic outcomes. Thus, adverse effects and drug interactions can be avoided, sudden cardiac death can be prevented, and the time between the commencement of drug treatment and the desired result can be shortened. This article presents a parametric model of the left ventricle automatically generated from patient-specific ultrasound images by applying an electromechanical model of the heart. Drug effects were prescribed through specific boundary conditions for inlet and outlet flow, ECG measurements, and calcium function for heart muscle properties. Genetic data from patients were incorporated through the material property of the ventricle wall. Apical view analysis involves segmenting the left ventricle using a previously trained U-net framework and calculating the bordering rectangle based on the length of the left ventricle in the diastolic and systolic cycle. M-mode view analysis includes bordering of the characteristic areas of the left ventricle in the M-mode view. After extracting the dimensions of the left ventricle, a finite elements mesh was generated based on mesh options, and a finite element analysis simulation was run with user-provided inlet and outlet velocities. Users can directly visualize on the platform various simulation results such as pressure-volume, pressure-strain, and myocardial work-time diagrams, as well as animations of different fields such as displacements, pressures, velocity, and shear stresses.

In Silico Clinical Trials for Cardiovascular Disease.

The SILICOFCM project mainly aims to develop a computational platform for in silico clinical trials of familial cardiomyopathies (FCMs). The unique characteristic of the platform is the integration of patient-specific biological, genetic, and clinical imaging data. The platform allows the testing and optimization of medical treatment to maximize positive therapeutic outcomes. Thus, adverse effects and drug interactions can be avoided, sudden cardiac death can be prevented, and the time between the commencement of drug treatment and the desired result can be shortened. This article presents a parametric model of the left ventricle automatically generated from patient-specific ultrasound images by applying an electromechanical model of the heart. Drug effects were prescribed through specific boundary conditions for inlet and outlet flow, ECG measurements, and calcium function for heart muscle properties. Genetic data from patients were incorporated through the material property of the ventricle wall. Apical view analysis involves segmenting the left ventricle using a previously trained U-net framework and calculating the bordering rectangle based on the length of the left ventricle in the diastolic and systolic cycle. M-mode view analysis includes bordering of the characteristic areas of the left ventricle in the M-mode view. After extracting the dimensions of the left ventricle, a finite elements mesh was generated based on mesh options, and a finite element analysis simulation was run with user-provided inlet and outlet velocities. Users can directly visualize on the platform various simulation results such as pressure-volume, pressure-strain, and myocardial work-time diagrams, as well as animations of different fields such as displacements, pressures, velocity, and shear stresses.

Multi-Omics Integrative Approach of Extracellular Vesicles: A Future Challenging Milestone.

In the era of multi-omic sciences, dogma on singular cause-effect in physio-pathological processes is overcome and system biology approaches have been providing new perspectives to see through. In this context, extracellular vesicles (EVs) are offering a new level of complexity, given their role in cellular communication and their activity as mediators of specific signals to target cells or tissues. Indeed, their heterogeneity in terms of content, function, origin and potentiality contribute to the cross-interaction of almost every molecular process occurring in a complex system. Such features make EVs proper biological systems being, therefore, optimal targets of omic sciences. Currently, most studies focus on dissecting EVs content in order to either characterize it or to explore its role in various pathogenic processes at transcriptomic, proteomic, metabolomic, lipidomic and genomic levels. Despite valuable results being provided by individual omic studies, the categorization of EVs biological data might represent a limit to be overcome. For this reason, a multi-omic integrative approach might contribute to explore EVs function, their tissue-specific origin and their potentiality. This review summarizes the state-of-the-art of EVs omic studies, addressing recent research on the integration of EVs multi-level biological data and challenging developments in EVs origin.

Integration of clinical and biological data into the non-invasive diagnostic management of HCC (LIRADS) algorithm: Contribution for LIRADS-4 nodules

Background: The Liver Imaging Reporting and Data System (LI-RADS) standardizes the interpretation and reporting of imaging examinations in patients at risk for Hepatocellular Carcinoma (HCC).

Predicting Gene-Drug-Disease Interactions by integrating Heterogeneous Biological Data Through a Network Model

Abstract Prediction of gene-drug-disease interactions have talented new insights in biology. Discovering unknown interactions will provide new therapeutic approaches to explore gene expressions. Recent improvements in machine learning techniques have gotten considerable interest due to higher efficiency, accurate results, and their lower cost. However, most of the studies were ignoring relevant associations, by representing only drug-disease interactions on a network while public available data offers a large variety of interactions. Additionally, some computational techniques used in this domain are faced with new challenges, related to the organization of heterogeneous data which suffer from a high imbalance rate since there are extensively more non-interacting gene-drug-disease triplets than interacting ones. In this paper we present integration of heterogeneous biological data about genes, drugs, and diseases to build a model, and building a new graph representation relating genedrug-disease interactions. Using extreme gradient boosting (XGBoost) algorithm, we have been able to extract a list of valid interactions about gene-drug-disease triplets, and a list of gene-drug pairs related to lung cancer. Keywords: Biological heterogeneous data, Data integration, Gene-DrugDisease interactions, Machine learning.