Biological Interpretation Research Articles

DiSMVC: a multi-view graph collaborative learning framework for measuring disease similarity.

Exploring potential associations between diseases can help in understanding pathological mechanisms of diseases and facilitating the discovery of candidate biomarkers and drug targets, thereby promoting disease diagnosis and treatment. Some computational methods have been proposed for measuring disease similarity. However, these methods describe diseases without considering their latent multi-molecule regulation and valuable supervision signal, resulting in limited biological interpretability and efficiency to capture association patterns. In this study, we propose a new computational method named DiSMVC. Different from existing predictors, DiSMVC designs a supervised graph collaborative framework to measure disease similarity. Multiple bio-entity associations related to genes and miRNAs are integrated via cross-view graph contrastive learning to extract informative disease representation, and then association pattern joint learning is implemented to compute disease similarity by incorporating phenotype-annotated disease associations. The experimental results show that DiSMVC can draw discriminative characteristics for disease pairs, and outperform other state-of-the-art methods. As a result, DiSMVC is a promising method for predicting disease associations with molecular interpretability. Datasets and source codes are available at https://github.com/Biohang/DiSMVC. Supplementary data are available at Bioinformatics online.

Tensor modeling of MRSA bacteremia cytokine and transcriptional patterns reveals coordinated, outcome-associated immunological programs

Abstract Methicillin-resistant Staphylococcus aureus (MRSA) bacteremia is a common and life-threatening infection that imposes up to 30% mortality even when appropriate therapy is used. Despite in vitro efficacy determined by minimum inhibitory concentration (MIC) breakpoints, antibiotics often fail to resolve these infections in vivo, resulting in persistent MRSA bacteremia. Recently, several genetic, epigenetic, and proteomic correlates of persistent outcomes have been identified. However, the extent to which single variables or their composite patterns operate as independent predictors of outcome or reflect shared underlying mechanisms of persistence is unknown. To explore this question, we employed a tensor-based integration of host transcriptional and cytokine datasets across a well-characterized cohort of patients with persistent or resolving MRSA bacteremia outcomes. This method yielded high correlative accuracy with outcomes and immunologic signatures united by transcriptomic and cytokine datasets. Results reveal that patients with persistent MRSA bacteremia exhibit signals of granulocyte dysfunction, suppressed antigen presentation, and deviated lymphocyte polarization. In contrast, patients with resolving bacteremia heterogeneously exhibit correlates of robust antigen-presenting cell trafficking and enhanced neutrophil maturation corresponding to appropriate T lymphocyte polarization and B lymphocyte response. These results suggest that transcriptional and cytokine correlates of persistent versus resolving bacteremia outcomes are complex and may not be disclosed by conventional modeling. In this respect, a tensor-based integration approach may help to reveal consensus molecular and cellular mechanisms and their biological interpretation.

Data-driven analysis of chemicals, proteins and pathways associated with peanut allergy: from molecular networking to biological interpretation

Peanut allergy is majorly related to severe food induced allergic reactions. Several food including cow’s milk, hen’s eggs, soy, wheat, peanuts, tree nuts (walnuts, hazelnuts, almonds, cashews, pecans and pistachios), fish and shellfish are responsible for more than 90 % of food allergies. Here, we provide promising insights using a large-scale data-driven analysis, comparing the mechanistic feature and biological relevance of different ingredients presents in peanuts, tree nuts (walnuts, almonds, cashews, pecans and pistachios) and soybean. Additionally, we have analysed the chemical compositions of peanuts in different processed form raw, boiled and dry-roasted. Using the data-driven approach we are able to generate new hypotheses to explain why nuclear receptors like the peroxisome proliferator-activated receptors (PPARs) and its isoform and their interaction with dietary lipids may have significant effect on allergic response. The results obtained from this study will direct future experimental and clinical studies to understand the role of dietary lipids and PPAR-isoforms to exert pro-inflammatory or anti-inflammatory functions on cells of the innate immunity and influence antigen presentation to the cells of the adaptive immunity.

Generation of genetic codes with 2-adic codon algebra and adaptive dynamics

This is a brief review on modeling genetic codes with the aid of 2-adic dynamical systems. In this model amino acids are encoded by the attractors of such dynamical systems. Each genetic code is coupled to the special class of 2-adic dynamics. We consider the discrete dynamical systems, These are the iterations of a function F:Z2→Z2, where Z2 is the ring of 2-adic numbers (2-adic tree). A genetic code is characterized by the set of attractors of a function belonging to the code generating functional class. The main mathematical problem is to reduce degeneration of dynamic representation and select the optimal generating function. Here optimality can be treated in many ways. One possibility is to consider the Lipschitz functions playing the crucial role in general theory of iterations. Then we minimize the Lip-constant. The main issue is to find the proper biological interpretation of code-functions. One can speculate that the evolution of the genetic codes can be described in information space of the nucleotide-strings endowed with ultrametric (treelike) geometry. A code-function is a fitness function; the solutions of the genetic code optimization problem are attractors of the code-function. We illustrate this approach by generation of the standard nuclear and (vertebrate) mitochondrial genetics codes.

Optimization and validation of metabolomics methods for feline urine and serum towards application in veterinary medicine

BackgroundMetabolomics is an emerging and powerful technology that offers a comprehensive view of an organism's physiological status. Although widely applied in human medicine, it is only recently making its introduction in veterinary medicine. As a result, validated metabolomics protocols in feline medicine are lacking at the moment. Since biological interpretation of metabolomics data can be misled by the extraction method used, species and matrix-specific optimized and validated metabolomic protocols are sorely needed. ResultsSystematic optimization was performed using fractional factorial experiments for both serum (n=57) and urine (n=24), evaluating dilution for both matrices, and aliquot and solvent volume, protein precipitation time and temperature for serum. For the targeted (n=76) and untargeted (n=1,949) validation of serum respectively, excellent instrumental, intra-assay and inter-day precision were observed (CV≤ 15% or 30%, respectively). Linearity deemed sufficient both targeted and untargeted (R2≥0.99 or 0.90, respectively). An appropriate targeted recovery between 70 and 130% was achieved. For the targeted (n=69) and untargeted (n=2,348) validation of the urinary protocol, excellent instrumental and intra-assay precision were obtained (CV≤ 15% or 30%, respectively). Subsequently, the discriminative ability of our metabolomics methods was confirmed for feline chronic kidney disease (CKD) by univariate statistics (n=41 significant metabolites for serum, and n=55 for urine, p-value<0.05) and validated OPLS-DA models (R2(Y)>0.95, Q2(Y)>0.65, p-value<0.001 for both matrices). SignificanceThis study is the first to present an optimized and validated wholistic metabolomics methods for feline serum and urine using ultra-high performance liquid chromatography coupled to quadrupole-Orbitrap high-resolution mass spectrometry. This robust methodology opens avenues for biomarker panel selection and a deeper understanding of feline CKD pathophysiology and other feline applications.

Identifying and avoiding radiation damage in macromolecular crystallography.

Radiation damage remains one of the major impediments to accurate structure solution in macromolecular crystallography. The artefacts of radiation damage can manifest as structural changes that result in incorrect biological interpretations being drawn from a model, they can reduce the resolution to which data can be collected and they can even prevent structure solution entirely. Inthis article, we discuss how to identify and mitigate against the effects of radiation damage at each stage in the macromolecular crystal structure-solution pipeline.

Association of whole-person eigen-polygenic risk scores with Alzheimer's disease.

Late-Onset Alzheimer's Disease (LOAD) is a heterogeneous neurodegenerative disorder with complex etiology and high heritability. Its multifactorial risk profile and large portions of unexplained heritability suggest the involvement of yet unidentified genetic risk factors. Here we describe the "whole person" genetic risk landscape of polygenic risk scores for 2218 traits in 2044 elderly individuals and test if novel eigen-PRSs derived from clustered subnetworks of single-trait PRSs can improve the prediction of LOAD diagnosis, rates of cognitive decline, and canonical LOAD neuropathology. Network analyses revealed distinct clusters of PRSs with clinical and biological interpretability. Novel eigen-PRSs (ePRS) from these clusters significantly improved LOAD-related phenotypes prediction over current state-of-the-art LOAD PRS models. Notably, an ePRS representing clusters of traits related to cholesterol levels was able to improve variance explained in a model of the brain-wide beta-amyloid burden by 1.7% (likelihood ratio test P = 9.02 × 10-7). All associations of ePRS with LOAD phenotypes were eliminated by the removal of APOE-proximal loci. However, our association analysis identified modules characterized by PRSs of high cholesterol and LOAD. We believe this is due to the influence of the APOE region from both PRSs. We found significantly higher mean SNP effects for LOAD in the intersecting APOE region SNPs. Combining genetic risk factors for vascular traits and dementia could improve current single-trait PRS models of LOAD, enhancing the use of PRS in risk stratification. Our results are catalogued for the scientific community, to aid in generating new hypotheses based on our maps of clustered PRSs and associations with LOAD-related phenotypes.

Haplotype function score improves biological interpretation and cross-ancestry polygenic prediction of human complex traits.

We propose a new framework for human genetic association studies: at each locus, a deep learning model (in this study, Sei) is used to calculate the functional genomic activity score for two haplotypes per individual. This score, defined as the Haplotype Function Score (HFS), replaces the original genotype in association studies. Applying the HFS framework to 14 complex traits in the UK Biobank, we identified 3619 independent HFS-trait associations with a significance of p < 5 × 10-8. Fine-mapping revealed 2699 causal associations, corresponding to a median increase of 63 causal findings per trait compared with single-nucleotide polymorphism (SNP)-based analysis. HFS-based enrichment analysis uncovered 727 pathway-trait associations and 153 tissue-trait associations with strong biological interpretability, including 'circadian pathway-chronotype' and 'arachidonic acid-intelligence'. Lastly, we applied least absolute shrinkage and selection operator (LASSO) regression to integrate HFS prediction score with SNP-based polygenic risk scores, which showed an improvement of 16.1-39.8% in cross-ancestry polygenic prediction. We concluded that HFS is a promising strategy for understanding the genetic basis of human complex traits.

Haplotype function score improves biological interpretation and cross-ancestry polygenic prediction of human complex traits

We propose a new framework for human genetic association studies: at each locus, a deep learning model (in this study, Sei) is used to calculate the functional genomic activity score for two haplotypes per individual. This score, defined as the Haplotype Function Score (HFS), replaces the original genotype in association studies. Applying the HFS framework to 14 complex traits in the UK Biobank, we identified 3619 independent HFS–trait associations with a significance of p &lt; 5 × 10−8. Fine-mapping revealed 2699 causal associations, corresponding to a median increase of 63 causal findings per trait compared with single-nucleotide polymorphism (SNP)-based analysis. HFS-based enrichment analysis uncovered 727 pathway–trait associations and 153 tissue–trait associations with strong biological interpretability, including ‘circadian pathway-chronotype’ and ‘arachidonic acid-intelligence’. Lastly, we applied least absolute shrinkage and selection operator (LASSO) regression to integrate HFS prediction score with SNP-based polygenic risk scores, which showed an improvement of 16.1–39.8% in cross-ancestry polygenic prediction. We concluded that HFS is a promising strategy for understanding the genetic basis of human complex traits.

Adjusting for covariates and assessing modeling fitness in machine learning using MUVR2.

Machine learning (ML) methods are frequently used in Omics research to examine associations between molecular data and for example exposures and health conditions. ML is also used for feature selection to facilitate biological interpretation. Our previous MUVR algorithm was shown to generate predictions and variable selections at state-of-the-art performance. However, a general framework for assessing modeling fitness is still lacking. In addition, enabling to adjust for covariates is a highly desired, but largely lacking trait in ML. We aimed to address these issues in the new MUVR2 framework. The MUVR2 algorithm was developed to include the regularized regression framework elastic net in addition to partial least squares and random forest modeling. Compared with other cross-validation strategies, MUVR2 consistently showed state-of-the-art performance, including variable selection, while minimizing overfitting. Testing on simulated and real-world data, we also showed that MUVR2 allows for the adjustment for covariates using elastic net modeling, but not using partial least squares or random forest. Algorithms, data, scripts, and a tutorial are open source under GPL-3 license and available in the MUVR2 R package at https://github.com/MetaboComp/MUVR2.

BootGSEA: a bootstrap and rank aggregation pipeline for multi-study and multi-omics enrichment analyses.

Introduction: Gene set enrichment analysis (GSEA) subsequent to differential expression analysis is a standard step in transcriptomics and proteomics data analysis. Although many tools for this step are available, the results are often difficult to reproduce because set annotations can change in the databases, that is, new features can be added or existing features can be removed. Finally, such changes in set compositions can have an impact on biological interpretation. Methods: We present bootGSEA, a novel computational pipeline, to study the robustness of GSEA. By repeating GSEA based on bootstrap samples, the variability and robustness of results can be studied. In our pipeline, not all genes or proteins are involved in the different bootstrap replicates of the analyses. Finally, we aggregate the ranks from the bootstrap replicates to obtain a score per gene set that shows whether it gains or loses evidence compared to the ranking of the standard GSEA. Rank aggregation is also used to combine GSEA results from different omics levels or from multiple independent studies at the same omics level. Results: By applying our approach to six independent cancer transcriptomics datasets, we showed that bootstrap GSEA can aid in the selection of more robust enriched gene sets. Additionally, we applied our approach to paired transcriptomics and proteomics data obtained from a mouse model of spinal muscular atrophy (SMA), a neurodegenerative and neurodevelopmental disease associated with multi-system involvement. After obtaining a robust ranking at both omics levels, both ranking lists were combined to aggregate the findings from the transcriptomics and proteomics results. Furthermore, we constructed the new R-package "bootGSEA," which implements the proposed methods and provides graphical views of the findings. Bootstrap-based GSEA was able in the example datasets to identify gene or protein sets that were less robust when the set composition changed during bootstrap analysis. Discussion: The rank aggregation step was useful for combining bootstrap results and making them comparable to the original findings on the single-omics level or for combining findings from multiple different omics levels.

Unlocking the potential: A novel prognostic index signature for acute myeloid leukemia

Acute myeloid leukemia (AML) is an aggressive malignancy characterized by challenges in treatment, including drug resistance and frequent relapse. Recent research highlights the crucial roles of tumor microenvironment (TME) in assisting tumor cell immune escape and promoting tumor aggressiveness. This study delves into the interplay between AML and TME. Through the exploration of potential driver genes, we constructed an AML prognostic index (AMLPI). Cross-platform data and multi-dimensional internal and external validations confirmed that the AMLPI outperforms existing models in terms of areas under the receiver operating characteristic curves, concordance index values, and net benefits. High AMLPIs in AML patients were indicative of unfavorable prognostic outcomes. Immune analyses revealed that the high-AMLPI samples exhibit higher expression of HLA-family genes and immune checkpoint genes (including PD1 and CTLA4), along with lower T cell infiltration and higher macrophage infiltration. Genetic variation analyses revealed that the high-AMLPI samples associate with adverse variation events, including TP53 mutations, secondary NPM1 co-mutations, and copy number deletions. Biological interpretation indicated that ALDH2 and SPATS2L contribute significantly to AML patient survival, and their abnormal expression correlates with DNA methylation at cg12142865 and cg11912272. Drug response analyses revealed that different AMLPI samples tend to have different clinical selections, with low-AMLPI samples being more likely to benefit from immunotherapy. Finally, to facilitate broader access to our findings, a user-friendly and publicly accessible webserver was established and available at http://bioinfor.imu.edu.cn/amlpi. This server provides tools including TME-related AML driver genes mining, AMLPI construction, multi-dimensional validations, AML patients risk assessment, and figures drawing.

702 Multimodal CT Imaging of Ischemic Stroke Thrombi Identifies Scale-invariant Radiomic Features That Reflect Clot Biology

INTRODUCTION: Biological interpretability of ischemic stroke clot imaging remains challenging. Paired CT/microCT imaging of ischemic stroke clots resected by thrombectomy has the potential to identify interpretable image features that are correlated among pre-treatment image modalities and post-treatment histopathology. METHODS: We performed multimodal CT-imaging and histology for 10 stroke clots retrieved by mechanical thrombectomy. Clots were manually segmented from co-registered, pre-treatment CTA and nCCT. For the same cases, resected clots were iodine-stained, and imaged with a ScanCo microCT-100 (4.9μm resolution). Afterwards, clots were subjected to histological processing (H&amp;E staining) and whole slide scanned (40X). Clot radiomic features(RFs) (n = 93 per modality, 279 total) were extracted using PyRadiomics and histological composition was computed using Orbit Image Analysis. Correlation analysis was used to test associations between microCT and CTA (or nCCT) RFs as well as between RFs and histological composition. Statistical significance was considered at R = 0.65 and q &lt; 0.05. RESULTS: From paired RF correlation analysis, we identified 23 scale-invariant RFs significantly correlated between microCT and CTA (18), and microCT and nCCT (5). Correlation of unpaired RFs identified 377 positively and 36 negatively correlated RFs between microCT and CTA, and 168 positively and 41 negatively correlated RFs between microCT and nCCT. Scale invariant RFs computed from CTA and nCCT demonstrated significant correlation with red blood cell and fibrin-platelet components, while microCT RFs were found to be correlated with WBC percent composition. CONCLUSIONS: This study is the first to complete a detailed statistical analysis of radiomic features across scale, and investigation of their relationship to histological composition. Our findings suggest that RFs from resected clot microCT can be used to identify strong and significant correlations between histology and pre-thrombectomy CT, with increased biological interpretability.

Development and validation of a deep learning-based model to predict response and survival of T790M mutant non-small cell lung cancer patients in early clinical phase trials using electronic medical record and pharmacokinetic data.

Epidermal growth factor receptor (EGFR) T790M mutation is the standard predictive biomarker for third-generation epidermal growth factor receptor tyrosine kinase inhibitor (EGFR-TKI) treatment. While not all T790M-positive patients respond to third-generation EGFR-TKIs and have a good prognosis, it necessitates novel tools to supplement EGFR genotype detection for predicting efficacy and stratifying EGFR-mutant patients with various prognoses. Mixture-of-experts (MoE) is designed to disassemble a large model into many small models. Meanwhile, it is also a model ensembling method that can better capture multiple patterns of intrinsic subgroups of enrolled patients. Therefore, the combination of MoE and Cox algorithm has the potential to predict efficacy and stratify survival in non-small cell lung cancer (NSCLC) patients with EGFR mutations. We utilized the electronic medical record (EMR) and pharmacokinetic parameters of 326 T790M-mutated NSCLC patients, including 283 patients treated with Abivertinib in phase I (n=177, for training) and II (n=106, for validation) clinical trials and an additional validation cohort 2 comprising 43 patients treated with BPI-7711. Furthermore, 18 patients underwent whole-exome sequencing for biological interpretation of CoxMoE. We evaluated the predictive performance for therapeutic response using the area under the curve (AUC) and the Concordance index (C-index) for progression-free survival (PFS). CoxMoE exhibited AUCs of 0.73-0.83 for predicting efficacy defined by best overall response (BoR) and achieved C-index values of 0.64-0.65 for PFS prediction in training and validating cohorts. The PFS of 198 patients with a low risk [median, 6.0 (range, 1.0-23.3) months in the abivertinib treated cohort; median 16.5 (range, 1.4-27.4) months in BPI-7711 treated cohort] of being non-responder increased by 43% [hazard ratio (HR), 0.56; 95% confidence interval (CI), 0.40-0.78; P=0.0013] and 50% (HR, 0; 95% CI, 0-0; P=0.01) compared to those at high-risk [median, 4.2 (range, 1.0-35) months in the abivertinib treated cohort; median, 11.0 (range, 1.4-25.1) months in BPI-7711 treated cohort]. Additionally, activated partial thromboplastin time (APTT), creatinine clearance (Ccr), monocyte, and steady-state plasma trough concentration utilited to construct model were found significantly associated with drug resistance and aggressive tumor pathways. A robust correlation was observed between APTT and Ccr with PFS (log-rank test; P<0.01) and treatment response (Wilcoxon test; P<0.05), respectively. CoxMoE offers a valuable approach for patient selection by forecasting therapeutic response and PFS utilizing laboratory tests and pharmacokinetic parameters in the setting of early-phase clinical trials. Simultaneously, CoxMoE could predict the efficacy of third-generation EGFR-TKI non-invasively for T790M-positive NSCLC patients, thereby complementing existing EGFR genotype detection.

Rescue of Mycobacterium bovis DNA Obtained from Cultured Samples during Official Surveillance of Animal TB: Key Steps for Robust Whole Genome Sequence Data Generation.

Epidemiological surveillance of animal tuberculosis (TB) based on whole genome sequencing (WGS) of Mycobacterium bovis has recently gained track due to its high resolution to identify infection sources, characterize the pathogen population structure, and facilitate contact tracing. However, the workflow from bacterial isolation to sequence data analysis has several technical challenges that may severely impact the power to understand the epidemiological scenario and inform outbreak response. While trying to use archived DNA from cultured samples obtained during routine official surveillance of animal TB in Portugal, we struggled against three major challenges: the low amount of M. bovis DNA obtained from routinely processed animal samples; the lack of purity of M. bovis DNA, i.e., high levels of contamination with DNA from other organisms; and the co-occurrence of more than one M. bovis strain per sample (within-host mixed infection). The loss of isolated genomes generates missed links in transmission chain reconstruction, hampering the biological and epidemiological interpretation of data as a whole. Upon identification of these challenges, we implemented an integrated solution framework based on whole genome amplification and a dedicated computational pipeline to minimize their effects and recover as many genomes as possible. With the approaches described herein, we were able to recover 62 out of 100 samples that would have otherwise been lost. Based on these results, we discuss adjustments that should be made in official and research laboratories to facilitate the sequential implementation of bacteriological culture, PCR, downstream genomics, and computational-based methods. All of this in a time frame supporting data-driven intervention.

Biological Testing and Interpretation of Laboratory Results Associated with Detecting Newborns with Substance Exposure.

Substance use during pregnancy is common, as is biological testing that is intended to help identify prenatal exposures. However, there is no standardized requirement for biological testing with either maternal or newborn specimens, nor is there standardization related to when testing occurs, how frequently testing occurs, what specimen(s) to test, what substances to test for, or how to perform testing. We review common specimen types tested to detect maternal and newborn substance exposure with a focus on urine, meconium, and umbilical cord tissue. We also review common analytical methods used to perform testing, including immunoassay, and mass spectrometry platforms. Considerations regarding the utilization of testing relative to the purpose of testing, the drug analyte(s) of interest, the specific testing employed, and the interpretation of results are emphasized to help guide decisions about clinical utilization of testing. We also highlight specific examples of unexpected results that can be used to guide interpretation and appropriate next steps. There are strengths and limitations associated with all approaches to detecting substance exposure in pregnant persons as well as biological testing to evaluate a newborn with possible substance exposure. Standardization is needed to better inform decisions surrounding evaluation of substance exposures in pregnant people and newborns. If biological sampling is pursued, testing options and results must be reviewed in clinical context, acknowledging that false-positive and -negative results can and do occur.

EpiSegMix: a flexible distribution hidden Markov model with duration modeling for chromatin state discovery.

Automated chromatin segmentation based on ChIP-seq (chromatin immunoprecipitation followed by sequencing) data reveals insights into the epigenetic regulation of chromatin accessibility. Existing segmentation methods are constrained by simplifying modeling assumptions, which may have a negative impact on the segmentation quality. We introduce EpiSegMix, a novel segmentation method based on a hidden Markov model with flexible read count distribution types and state duration modeling, allowing for a more flexible modeling of both histone signals and segment lengths. In a comparison with existing tools, ChromHMM, Segway, and EpiCSeg, we show that EpiSegMix is more predictive of cell biology, such as gene expression. Its flexible framework enables it to fit an accurate probabilistic model, which has the potential to increase the biological interpretability of chromatin states. Source code: https://gitlab.com/rahmannlab/episegmix.

DeepKEGG: a multi-omics data integration framework with biological insights for cancer recurrence prediction and biomarker discovery.

Deep learning-based multi-omics data integration methods have the capability to reveal the mechanisms of cancer development, discover cancer biomarkers and identify pathogenic targets. However, current methods ignore the potential correlations between samples in integrating multi-omics data. In addition, providing accurate biological explanations still poses significant challenges due to the complexity of deep learning models. Therefore, there is an urgent need for a deep learning-based multi-omics integration method to explore the potential correlations between samples and provide model interpretability. Herein, we propose a novel interpretable multi-omics data integration method (DeepKEGG) for cancer recurrence prediction and biomarker discovery. In DeepKEGG, a biological hierarchical module is designed for local connections of neuron nodes and model interpretability based on the biological relationship between genes/miRNAs and pathways. In addition, a pathway self-attention module is constructed to explore the correlation between different samples and generate the potential pathway feature representation for enhancing the prediction performance of the model. Lastly, an attribution-based feature importance calculation method is utilized to discover biomarkers related to cancer recurrence and provide a biological interpretation of the model. Experimental results demonstrate that DeepKEGG outperforms other state-of-the-art methods in 5-fold cross validation. Furthermore, case studies also indicate that DeepKEGG serves as an effective tool for biomarker discovery. The code is available at https://github.com/lanbiolab/DeepKEGG.

Abstract 7373: Generative Bayesian networks for augmentation of molecular data from commercial genetics panels

Abstract Background: Computational models in medical research which use molecular features to predict patient sensitivities or outcomes are traditionally limited to producing a scalar output for a given disease phenotype, e.g. progression-free survival or drug response. Even for relatively intelligible models, it is generally difficult to place confidence limits on the prediction or develop an intuition for how a prediction varies with respect to “what-if” changes in those features or how sensitive a prediction is to changes in input variables. Moreover, while essential for understanding and modeling tumor behavior, molecular features from patient tumors in real-world settings are typically limited to profiles of hundreds of genes, while the biologically relevant inputs to the patient’s condition can be much larger. Approach: To generate data required to enhance the interpretation of predictive models and extend the utility of real-world patient oncology datasets, a distribution of conditions consistent with the observed features and their incidence across cancer patient cohorts is required. We hypothesized we could synthesize patient samples drawn from the joint probability distribution of a broader universe of features when a subset of them is held fixed at the observed values. We model this joint distribution by learning a Bayesian network over a broad feature set for which some training data is available and generate feature profiles by applying Gibbs sampling to the learned network. Results: We assessed the potential clinical utility of using these generated feature profiles in predicting drug response and enhancing biological interpretation of different model outputs. We found generative somatic mutations were useful for predicting cancer patient outcomes, including drug response predictions in breast cancer tumors from AACR Project GENIE. Synthesized patient feature profiles enhanced biological interpretability of nominal panel data (100-200 genes), providing a means to assign probabilities to outcomes and uncovering previously described mechanisms of drug response in real-world patient data. Conclusion: We introduce an approach that leverages Bayesian networks to synthesize richly annotated patient feature profiles from limited molecular data routinely collected in real-world settings. This approach addresses challenges associated with limited molecular data, biological interpretability and evaluation of predictive models and enhances the utility of real-world datasets for cancer research. Citation Format: Dillon H. Tracy, Jeff Sherman, Maayan Baron. Generative Bayesian networks for augmentation of molecular data from commercial genetics panels [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7373.

Abstract 3546: Putting genes in context by linking gene signatures to established cancer hallmarks

Abstract The concept of the "Hallmarks of Cancer" provides a valuable framework for understanding fundamental organizing principles common to various types of cancers. However, the absence of a consensus gene set for cancer hallmarks poses data comparison and integration challenges, resulting in diverse biological interpretations across studies. Here, we first established a consensus cancer hallmark-gene set by merging data from available mapping resources. We searched the scientific literature for relevant publications from which a list of hallmark genes could be reconstructed. By consolidating data from seven publications, we identified 6,763 candidate cancer hallmark genes associated with ten cancer hallmarks. Of these, 1574 genes were part of at least two resources, forming a "core" hallmark gene set. Then, we compared the enrichment of cancer hallmarks by analyzing the prognostic genes associated with overall survival across twelve solid tumor types. Notably, the hallmark "tissue invasion and metastasis" was most prominent in the stomach (p=1.2E-11), pancreatic (p=1.5E-09), bladder (p=1.8E-08), and ovarian cancers (p=0.0002), aligning with their heightened potential to spread. "Sustained angiogenesis" predominated in lung squamous carcinomas (p=4.5E-08), while "genome instability" showed strong enrichment in lung adenocarcinomas (p=1.3E-09), liver (p=1.4E-10), pancreatic (p=1E-5) and kidney cancers (p=0.0025). Compelling evidence links lifestyle choices like alcohol consumption and smoking to genomic instability across tumors, suggesting a potential common underlying mechanism of tumorigenesis. Pancreatic cancers displayed the highest enrichment of hallmarks (8 out of 10), emphasizing the disease's complexity, while in melanomas, liver, and kidney cancer, a single hallmark (,,genomic instability”) was enriched among the prognostic markers of survival. These findings underscore the utility of the hallmark concept as an effective organizational tool, particularly when establishing a clear connection between genes and biological functions. Finally, we integrated the data into a publicly accessible and user-friendly online tool (available at www.cancerhallmarks.com) that allows users to identify the most relevant cancer-associated "candidate hallmark genes" in their datasets. In summary, we have established a consensus list of genes associated with cancer hallmarks. The analysis of survival-associated genes revealed a unique pattern of hallmark enrichment for each tumor type. Citation Format: Otília Menyhart, William Kothalawala, Balázs Győrffy. Putting genes in context by linking gene signatures to established cancer hallmarks [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3546.