Why Liquid Biopsy Multiomics?
Advances in next-generation sequencing (NGS) technologies have enabled blood-based detection of small quantities of circulating nucleic acids with high specificity. Circulating tumor DNA (ctDNA) liquid biopsies (LBx) are now a reliable tool for obtaining valuable insights into tumor-specific genomic alterations, identifying patients eligible for targeted therapies, monitoring tumor dynamics, and serving as endpoints in FDA clinical trials. However, there remains a significant unmet need for LBx technologies that capture heterogeneity of critical host immune responses that are required for effective therapeutic response.
To bridge this gap, we developed LITOSeek™, the Liquid ImmunoTranscriptomic profiling platform, combining NGS and AI technologies to analyze blood RNA biomarkers that provide a deeper understanding of systemic immunity, and enable predictions of patient responses to therapy. This multiomics LBx platform integrates genomic mutations (ctDNA) and whole-blood transcriptomics data to generate robust, patient-specific insights with predictive capabilities.
LITOSeek™ not only enhances early disease detection but also illuminates each patient’s immune landscape, predicts treatment response, and accelerates development of novel therapeutics. The platform addresses the unmet need for a comprehensive liquid biopsy multiomics solution, offering transformative potential in improving clinical diagnostics and biopharma drug development.
Unmet needs in RNA liquid biopsy testing:
1. Enhanced Sensitivity for Early Detection of Disease & Treatment Response:
Blood-based RNA biomarkers provide a dynamic and comprehensive view of host immune responses and have demonstrated over a three-fold increase in sensitivity for detecting pre-cancerous colorectal lesions (advanced adenomas) comparedto traditional ctDNA methods, making it a valuable tool for early disease detection, and for detection of Immuno-Pharmaco Dynamic biomarkers of response to therapy.
2. Dynamic Monitoring of Tumor-Host Interaction:
ImmunoTranscriptomic profiling tracks changes in gene expression patterns over time capturing real-time immune dynamics and providing a deep understanding of the systemic host immune response to the evolution of the disease. Multimomic combination of LBx RNA and LBx ctDNA provides a comprehensive view on the evolution of tumor-host interactions with Precise and Predictive Patient Profiling that provide early prediction of clinical benefit, and support optimization of therapeutic intervention.
3. Insights Into Mechanisms of Resistance and Therapy Development:
ImmunoTranscriptomic profiling enables detailed analysis of immune- related markers and pathways, offering valuable insights into mechanisms of resistance in patient sub-populations that do not respond to therapeutic intervention. These capabilities are transformative for therapy development in clinical trials, paving the way for personalized and effective treatment strategies..
How it works
Liquid Biopsy Multiomics Workflow
The AI-enabled LITOSeek™ platform has been developed to overcome challenges in whole blood RNA analysis, with Quality Control (QC) protocols that ensure the highest levels of precision, robustness, and reproducibility. The platform is compatible with other existing technologies for ctDNA and proteomics profiling and provides multiomics assessment of data readouts with unprecedented predictive capabilities.
Blood RNA Data Generation
Novigenix has established standardized protocols under ISO-13485 certification for extraction of RNA from stabilized whole blood to conduct transcriptome-wide RNA sequencing. The protocol guarantees accuracy and consistency in results by incorporating a number of QC strategies such as spike-in standards during sequencing, enabling reliable and reproducible ImmunoTranscriptomic profiling analysis.
Tracking & Quantifying Technical Noise in Blood Samples
Identifying biomarkers in blood requires exceptional precision, and the quality of collected samples significantly influences downstream analyses and results.
Drawing from extensive experience in analyzing thousands of RNA-seq blood samples, LITOSeek™ protocols (Figure 1) emphasize the critical importance of identifying low-quality samples as they can introduce inaccuracies and compromise the robustness of results. Traditional tools fail to capture the combined effects of technical variables, making reliable sample quality assessment challenging. To address this, LITOSeek™ employs a rigorous approach that traces parameters affecting sample processing through RNA extraction and sequencing. NoviQC ensures sample quality by leveraging advanced machine learning models that detect complex, nonlinear quality patterns, and provide quantitative quality control probability scores that enhance the reliability of biomarker discovery and enables targeted evaluation of marginal patient samples when necessary.
Data Cleaning: Noise Filtering & Harmonization
Denoising RNA-seq Data: Blood RNA-seq data is prone to noise due to the influence of numerous biological and technical parameters that compromise the accuracy of gene expression data. It is therefore imperative to distinguish noisy genes from true biological signals and ensure the reliability of identified biomarkers. LITOSeek™ incorporates a proprietary state-of-the-art algorithm, NoviDeNoise, designed to address this challenge by effectively filtering out unstable genes using advanced mathematical methods. This ensures that identified biomarkers accurately reflect underlying biological activity rather than artifacts caused by technical variability.
NoviDeNoise’s robustness stems from an extensive experimental framework, developed with over 500 proprietary blood samples, each represented by up to 12 technical replicates under diverse conditions. These conditions include baseline and post-stimulation states, various sequencing protocols, and sequencing conducted at multiple facilities. The meticulous design of NoviDeNoise ensures the algorithm’s ability to handle a wide range of scenarios, consistently delivering reliable results despite technical and biological variations in the data. This rigorous framework ensures consistent gene expression selection across diverse sequencing protocols and significantly increases robustness and reproducibility of results (Figure 2).
Data Integration and Harmonization with NoviMetaHarmony and NoviCountHarmony
Data Integration and Harmonization with NoviMetaHarmony and NoviCountHarmony Harmonization is the process of integrating diverse datasets into a unified framework, ensuring compatibility and consistency for robust downstream analyses. LITOSeek™ addresses this challenge on two levels: harmonizing clinical and technical metadata, and aligning transcriptomics data. This is achieved through development of two advanced pipelines:
NoviCountHarmony
Designed to harmonize RNA-seq data, this pipeline employs a mathematical approach to remove batch effects arising from sequencing runs, while preserving biological signals. NoviCountHarmony facilitates integration of sequencing batches enabling comprehensive analyses of small and large batches of patient samples across stages of clinical development.
NoviMetaHarmony
This pipeline leverages Large Language Models (LLMs) to standardize over 500 clinical and metadata that describe each sample. By ensuring consistency in both variable names and values across datasets, NoviMetaHarmony streamlines metadata integration for AI-enabled cross-cohort analyses. Together, these pipelines ensure that both metadata and transcriptomics data are integrated reliably, empowering accurate and reproducible insights from diverse datasets.
LITOSeek™ Knowledge Base (KB)
LITOSeek™ KB serves as the foundational component of the LITOSeek™ platform, functioning as a repository of clinically-annotated, RNA-seq datasets derived from proprietary patient blood samples. As a centralized SQL-based database, LITOSeek™ KB integrates variables associated with patient clinical samples and RNA-seq data. The distinctive design of this database, coupled with its integrated data harmonization tools enables LITOSeek™ to incorporate structured and cleaned datasets crucial for the development of precise and predictive models.
Data Analytics
LITOSeek™ Platform Analytical Workflow
NoviScore: Biomarker Selection.
At the heart of the LITOSeek™ platform is the proprietary feature selection pipeline NoviScore that integrates both Univariate and Multivariate algorithm classes to identify and comprehensively evaluate potential biomarkers. The essence of NoviScore lies in its ability to judiciously analyze gene sets from various angles, ensuring that all viable biomarkers are captured for scoring and integration, effectively prioritizing true signals, and diminishing the influence of misleading artifacts.
NoviContext: Biomarker Biological Interpretation & Contextualization.
In the advanced stage of building a machine learning model, the highest NoviScore-ranked genes are processed through the NoviContext pipeline by our team of immunology and immuno-oncology experts who interpret the results by delving deep into the biological significance of the identified biomarkers. This crucial step involves a comprehensive post-differential expression analysis, utilizing LITOSeek™ automated processes that integrate diverse methods and databases collected samples significantly influences downstream analyses and results.
NoviSig and NoviBoost: ML-based Feature Reduction, Model Development & Evaluation.
At the core of our machine learning model development lies the process of feature reduction and model development with NoviSig and NoviBoost pipelines. Following rigorous quality control, cleaning, structuring, and harmonizing of RNA- seq data alongside patient metadata, the data are housed within a specially engineered database optimized for managing
substantial data volumes, where NoviSig and NoviBoost work in tandem to craft reliable and clinically meaningful predictive models. The combination of high-quality data, cutting-edge machine learning techniques, and sophisticated database infrastructure serves as a robust foundation for building highly accurate models for Precise & Predictive Patient Profiling.