Drug analysis and repurposing platform with synergy and disease clustering capabilities

Abstract

A method for maintaining a gene description dataset associated with a plurality of genes with respect to a given disease, wherein the gene description dataset is a human- comprehensible text-based dataset generated by a Large Language Model (LLM); maintaining a dataset of drug mechanisms of action (MOAs) associated with a plurality of drugs; embedding the gene descriptions and the MOAs into high-dimensional vectors using a text embedding technique to generate gene description vectors and MOA vectors; calculating distances between the gene description vectors and the MOA vectors to generate distance scores; and ranking the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

Description

Docket No.: 063697-505001WQDRUG ANALYSIS AND REPURPOSING PLATFORM WITH SYNERGY AND DISEASE CLUSTERING CAPABILITIESCROSS-REFERENCE

[0001] This application claims the benefit of U.S. Provisional Application No. 63 / 720,065 filed November 13, 2024, which application is incorporated herein by reference.TECHNICAL FIELD

[0002] The subject matter described herein relates to drug discovery and biomedical data analysis technology, specifically platforms, systems and methods for drug repurposing, predicting drug approval likelihood, synergy prediction, and disease cluster generation with associated cluster-specific drug ranking, using advanced data processing and vector-based analysis techniques.BACKGROUND

[0003] In the field of biomedical research and drug discovery, identifying effective drugs for repurposing and predicting their potential approval likelihood are crucial for accelerating treatment options and optimizing research and drug development investments. Researchers often face challenges in assessing drug suitability for new indications or patient subgroups and discovering combinations that can yield synergistic effects. Conventional methods tend to rely heavily on isolated data points and singular biological interactions, which may limit the scope of discovery and overlook broader, systemic relationships within complex biological data. This can result in missed opportunities for repurposing drugs efficiently or finding promising drug synergies.Docket No.: 063697-505001WQ

[0004] Existing approaches to drug repurposing and synergy prediction are often constrained by traditional analysis models that focus on specific targets or pathways without integrating comprehensive knowledge across different data sources. These methods can be fragmented and lack the capability to synthesize multi-dimensional biological data into meaningful patterns that aid in identifying potential new uses for existing drugs. Additionally, the absence of robust models for analyzing the potential effectiveness of drugs in specific patient subgroups or for generating disease clusters further limits the ability to make precise and personalized treatment decisions.

[0005] There is a need for innovative systems capable of synthesizing diverse biological data into cohesive insights that facilitate drug repurposing, predict approval likelihood for drugs under development, identify synergistic drug combinations, and enable disease cluster generation. Such systems would empower researchers to utilize data more effectively, uncover hidden relationships between drugs and diseases, and support targeted treatment strategies tailored to specific patient subgroups, ultimately contributing to more efficient and adaptive approaches in drug development and healthcare.SUMMARY

[0006] The present disclosure relates to a computer-implemented method, a computer program product, and a system for ranking drugs based on their mechanisms of action (MO As) relative to gene descriptions for a given disease. The invention utilizes a Large Language Model (LLM) to generate human-comprehensible gene descriptions and embeds them into high-dimensional vectors to evaluate their proximity to MOA vectors, aiding in drug ranking and selection.

[0007] In one aspect, a computer-implemented method is provided that involves maintaining a gene description dataset associated with a plurality of genes for a given disease.Docket No.: 063697-505001WO The gene description dataset is a human-comprehensible text-based dataset generated by an LLM. A separate dataset of drug MO As associated with a plurality of drugs is also maintained. The method includes embedding both the gene descriptions and the MOAs into highdimensional vectors using a text embedding technique to create gene description vectors and MOA vectors. Distances between the gene description vectors and the MOA vectors are calculated to generate distance scores. These scores are then used to rank the drugs based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

[0008] In one variation of the method, the gene description dataset is generated by identifying a plurality of genes associated with the given disease. A scraping process collects descriptions of each gene from multiple sources, ensuring comprehensive coverage. A standardized query prompt is designed and applied to extract descriptions for each gene, specifying consistent context depth and length. To focus on the generic attributes of the genes, references to the given disease are removed from the descriptions.

[0009] In one variation, the method further incorporates a text embedding technique that applies a language model to convert the gene descriptions and MOA texts into highdimensional vector representations. This embedding technique ensures that the relationships between the textual data are preserved in the vector space, facilitating the comparison between genes and drug mechanisms.

[0010] In one variation, the calculation of distances between vectors may use one or more distance metrics, such as cosine similarity, Euclidean distance, Manhattan distance, or Minkowski distance, to generate distance scores that indicate the similarity between gene description vectors and MOA vectors.Docket No.: 063697-505001WO

[0011] In one variation, the method further includes aggregating the distance scores by calculating an average or weighted average of the distances between the gene description vectors and the MOA vectors. This aggregation helps determine the overall suitability of each drug for treating the given disease.

[0012] In one variation, the method refines the ranking by identifying drug combinations that maximize the number of genes associated with the given disease that have distance scores below a predetermined threshold. This ensures that combinations that may target a broader set of genes are considered in the ranking process.

[0013] In one variation, the method includes identifying subgroups of patients for whom a particular drug may be more effective by analyzing the distribution of distance scores and the expression variability of genes related to distinct phenotypes of the disease. This analysis aids in personalized treatment strategies by identifying drugs that align more closely with specific patient subgroups.

[0014] In one aspect, a computer program product is provided, wherein the computer program product comprises a non-transient machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising: maintaining a gene description dataset associated with a plurality of genes for a given disease. The gene description dataset is a human-comprehensible text-based dataset generated by an LLM. A separate dataset of drug MO As associated with a plurality of drugs is also maintained. The method includes embedding both the gene descriptions and the MO As into high-dimensional vectors using a text embedding technique to create gene description vectors and MOA vectors. Distances between the gene description vectors and the MOA vectors are calculated to generate distance scores. These scores are then used to rank the drugs based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.Docket No.: 063697-505001WQ

[0015] In one variation, for the computer program product, the operations include generating the gene description dataset by identifying genes associated with a given disease, scraping data sources for gene descriptions, applying standardized prompts to extract these descriptions, and removing disease references to create the dataset. This ensures that the generated dataset remains general yet relevant for use with various diseases.

[0016] In one variation, the program product utilizes a text embedding technique, applying a language model to convert both gene descriptions and MOA texts into vector representations.

[0017] In one variation, the calculation of distances between vectors within the computer program product may employ distance metrics, including cosine similarity, Euclidean distance, Manhattan distance, or Minkowski distance, to ensure flexibility in determining the most relevant metric for different contexts.

[0018] In one variation, the operations may further include aggregating the distance scores by computing an average or weighted average to assess the overall drug suitability for the given disease.

[0019] In one variation, the computer program product may include operations for identifying drug combinations that maximize the number of relevant genes with favorable distance scores, thus supporting multi-target drug strategies.

[0020] The computer program product may also analyze the distribution of distance scores and the variability in gene expression across different disease phenotypes to identify patient subgroups that may respond better to specific drugs, enhancing the personalized treatment approach.

[0021] In another aspect, the present disclosure relates to a system comprising a programmable processor and a non-transient machine-readable medium storing instructions.Docket No.: 063697-505001WQ The system performs operations that include maintaining a gene description dataset associated with a plurality of genes with respect to a given disease, wherein the gene description dataset is a human-comprehensible text-based dataset generated by a Large Language Model (LLM); maintaining a dataset of drug mechanisms of action (MOAs) associated with a plurality of drugs; embedding the gene descriptions and the MOAs into high-dimensional vectors using a text embedding technique to generate gene description vectors and MOA vectors; calculating distances between the gene description vectors and the MOA vectors to generate distance scores; and ranking the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

[0022] In one variation, the system may further involve generating the gene description dataset by identifying relevant genes, scraping multiple data sources for descriptions, using a standardized prompt for consistency, and removing references to the disease to refine the dataset.

[0023] In one variation, the system applies a language model for embedding the textual data into vector representations, preserving the semantic meaning of the data in a highdimensional space. This facilitates accurate distance calculations between the vectors.

[0024] In one variation, the system calculates distances between the vectors using metrics such as cosine similarity, Euclidean distance, Manhattan distance, or Minkowski distance. This flexibility in distance metrics ensures that the most effective measure is applied for each use case.

[0025] In one variation, the system supports the identification of drug combinations that target a broad range of genes with favorable distance scores, aiding in comprehensive treatment strategies for the given diseaseDocket No.: 063697-505001WO

[0026] Implementations of the current subject matter can include, but are not limited to, methods consistent with the descriptions provided herein as well as articles that include a tangibly embodied machine-readable medium operable to cause one or more machines (e.g., computers, etc.) to result in operations implementing one or more of the described features. Similarly, computer systems are also described that may include one or more processors and one or more memories coupled to the one or more processors. A memory, which can include a computer-readable storage medium, may include, encode, store, or the like one or more programs that cause one or more processors to perform one or more of the operations described herein. Computer implemented methods consistent with one or more implementations of the current subject matter can be implemented by one or more data processors residing in a single computing system or multiple computing systems. Such multiple computing systems can be connected and can exchange data and / or commands or other instructions or the like via one or more connections, including but not limited to a connection over a network (e.g. the Internet, a wireless wide area network, a local area network, a wide area network, a wired network, or the like), via a direct connection between one or more of the multiple computing systems, etc.

[0027] The details of one or more variations of the subject matter described herein are set forth in the accompanying drawings and the description below. Other features and advantages of the subject matter described herein will be apparent from the description and drawings, and from the claims. The claims that follow this disclosure are intended to define the scope of the protected subject matter.DESCRIPTION OF DRAWINGS

[0028] The accompanying drawings, which are incorporated in and constitute a part of this specification, show certain aspects of the subject matter disclosed herein and, togetherDocket No.: 063697-505001WG with the description, help explain some of the principles associated with the disclosed implementations. In the drawings,

[0029] FIG. 1 is a diagram illustrating an example of the platform, in accordance with one or more embodiments of the current subject matter.

[0030] FIG. 2 is a diagram illustrating an example of the gene description dataset module, in accordance with one or more embodiments of the current subject matter.

[0031] FIG. 3 is a diagram illustrating an example of the text embedding engine, in accordance with one or more embodiments of the current subject matter.

[0032] FIG. 4 is a diagram illustrating a flow chart of a process for ranking drugs for repurposing to treat a given disease, in accordance with one or more embodiments of the current subject matter.

[0033] FIG. 5 depicts a block diagram illustrating a computing system consistent with implementations of the current subject matter.

[0034] When practical, like labels are used to refer to same or similar items in the drawings.DETAILED DESCRIPTION

[0035] The details of one or more variations of the subject matter described herein are set forth in the accompanying drawings.

[0036] To overcome existing challenges in drug discovery and analysis, the present disclosure introduces an integrated platform designed for efficient data processing and analysis. This platform leverages multiple interconnected modules to enable drug repurposing, synergy prediction, and the identification of patient subgroups. The following sections provide a detailed explanation of the platform’s architecture and functionality, showcasing how itsDocket No.: 063697-505001WG components collaborate to achieve robust and insightful results in drug analysis and disease research.

[0037] FIG. 1 is a diagram illustrating an example of the platform 100, in accordance with one or more embodiments of the current subject matter. In some embodiments, as shown in FIG. 1, the platform 100 may comprise multiple interconnected components, each serving a unique function within the overall system. The components may include the gene description dataset module 102, MO A (mechanisms of action) dataset module 104, text embedding engine 106, distance calculation module 108, and ranking and analysis module 110. These components collaborate seamlessly to facilitate drug repurposing, synergy prediction, and patient subgroup identification.

[0038] In some embodiments, the gene description dataset module 102 is responsible for maintaining a comprehensive dataset of gene descriptions associated with specific diseases. The gene description dataset module 102 may gather data from various sources, such as scientific literature, genomic databases, proteomic databases, metabolomic databases, and clinical records, to compile a diverse and complete set of descriptions. The module may process the collected data to maintain consistency and store it in a structured format, allowing for efficient access and integration. In some embodiments, the dataset is human-comprehensible text-based dataset. The gene description dataset module 102 serves as the initial source of information that feeds into subsequent components. For example, the gene description dataset module 102 may interact directly with the text embedding engine 106, providing the textual data needed for embedding into high-dimensional vectors. This interaction ensures a continuous flow of prepared data for downstream analysis. In some embodiments, the platform utilizes a Large Language Model (LLM), such as a transformer-based architecture trained on biomedical corpora, to generate human-comprehensible gene descriptions. The LLM processes input prompts designed to extract relevant biological information and outputs detailedDocket No.: 063697-505001WO descriptions that capture the functional and structural attributes of genes. This LLM-generated text is structured to maintain consistency and ensure comprehensibility, allowing downstream embedding techniques to accurately represent the gene information in vector form. The LLM leverages its training on extensive biomedical literature to produce outputs that align with domain-specific terminology and context, facilitating meaningful comparisons with drug mechanisms of action.

[0039] The MO A dataset module 104 may manage and maintain a dataset of drug mechanisms of action. This dataset may include detailed descriptions of how different drugs interact with biological systems to exert their effects, derived from sources such as clinical trial results, pharmacological studies, and regulatory filings. The MO A dataset module 104 may use automated data verification mechanisms to confirm the accuracy and currency of the data it stores. In some embodiments, the MOA dataset module 104 may work in tandem with the gene description dataset module 102 by providing complementary data to the text embedding engine 106 for parallel processing, enabling comprehensive comparative analysis.

[0040] The text embedding engine 106 may utilize advanced language models (e.g., Large Language Model (LLM)) to convert the gene descriptions and MOA data into highdimensional vector representations. These models, for example, may include transformerbased architectures or embeddings pre-trained on biomedical corpora, which enhance the representation of complex biological text. The text embedding engine 106 transforms input data from the gene description dataset module 102 and the MOA dataset module 104 into a quantitative format that can be used for further analysis. In some embodiments, the text embedding engine 106 may process data via a data processing pipeline to generate vector outputs, which are then passed to the distance calculation module 108. In some embodiments, the text embedding engine 106 applies advanced embedding techniques to convert both gene descriptions and MOA data into high-dimensional vector representations. These techniquesDocket No.: 063697-505001WG may include transformer-based language models such as BERT (Bidirectional Encoder Representations from Transformers) or GPT (Generative Pre-trained Transformer), which are trained on domain-specific biomedical corpora to enhance contextual understanding. The embedding process involves tokenizing input text, encoding the tokens into contextualized vector representations, and normalizing these vectors to ensure consistency across datasets. By using these models, the platform maintains semantic integrity, allowing for accurate comparisons between gene and MOA vectors in a high-dimensional space. This embedding approach supports the analysis of complex relationships in biological data, enabling downstream applications such as distance calculations and drug ranking.

[0041] The distance calculation module 108 may compute the distances between the gene description vectors and the MOA vectors using various distance metrics, such as cosine similarity, Euclidean distance, Manhattan distance, and Minkowski distance. The distance calculation module 108 assesses the relationships between genes and drug mechanisms by measuring the proximity of these vectors in high-dimensional space. In some embodiments, the distance calculation module 108 may be configured to use optimized algorithms for faster computation, enhancing the platform’s ability to process large datasets efficiently. The module may receive input from the text embedding engine 106 and output distance scores, which are essential for determining drug relevance and potential efficacy.

[0042] The ranking and analysis module 110 may aggregate the distance scores generated by the distance calculation module 108 and rank the drugs based on their suitability for treating a given disease. The ranking and analysis module 110 may employ methods for combining distance scores through weighted averages or other statistical techniques to refine drug ranking. Additionally, the module may identify subgroups of patients who may respond more effectively to specific drugs by analyzing the distribution of distance scores in conjunction with the expression variability of genes associated with distinct phenotypes of theDocket No.: 063697-505001WQ given disease. This analysis enables the identification of variability patterns and correlations that inform the selection of drugs tailored to specific patient profiles. By examining these distribution patterns, the method supports personalized treatment strategies that align with the unique genetic expression characteristics of patient subgroups. The outputs from the ranking and analysis module 110 may further include potential drug combinations that enhance therapeutic coverage across genes with favorable distance scores. These insights may be leveraged to inform drug repurposing or refocusing strategies and clinical decision-making, creating a robust tool for targeted therapeutic exploration within platform 100. In some embodiments, the ranking and analysis module 110 may aggregate distance scores by calculating an average or weighted average of the distances between gene description vectors and MOA vectors. This aggregation can be implemented using statistical techniques where weights are assigned based on the relative importance of each gene or its associated distance score, allowing for a more refined and contextually relevant drug ranking. The weighted average approach enables the system to prioritize specific genes that may have higher relevance to the disease, thus providing a more nuanced assessment of drug suitability. By incorporating various aggregation methods, the platform can adapt to different analytical needs and enhance the accuracy of drug ranking outcomes.

[0043] In some embodiments, the ranking and analysis module 110 implements a weighted averaging technique to combine distance scores effectively. Weights may be assigned based on predefined criteria, such as gene relevance scores or clinical importance, to refine the aggregate score for ranking drugs. This method enables more context-aware ranking, providing an enhanced evaluation that highlights drugs with stronger alignment to the disease-relevant gene profile. Additionally, statistical measures like variance and confidence intervals may be used to assess the robustness of the aggregated scores.Docket No.: 063697-505001WG

[0044] In some embodiments, the ranking and analysis module 110 may be configured to identify subgroups of patients who may respond more effectively to specific drugs by analyzing the distribution of distance scores in relation to the expression variability of genes associated with distinct phenotypes of the given disease. This process involves evaluating patterns within the calculated distance scores to reveal correlations between drug MO As and specific gene expression profiles. By incorporating gene expression variability analysis, the platform can distinguish patient subgroups characterized by different phenotypic responses, enabling a more tailored approach to treatment selection and drug repurposing. This analysis supports personalized medicine strategies by matching therapeutic options to subgroups with aligned gene-MOA interaction profiles.

[0045] In some embodiments, the platform incorporates statistical and machine learning algorithms to identify drug combinations and patient subgroups effectively. For drug combinations, the ranking and analysis module 110 may use optimization algorithms that evaluate combinations based on the aggregated distance scores and their ability to target a wider array of relevant genes with favorable scores. For identifying patient subgroups, the platform may apply clustering algorithms such as k-means, hierarchical clustering, or principal component analysis (PCA) to group patients based on the distribution of distance scores and gene expression variability. These algorithms analyze patterns in the data to segment patients into subgroups characterized by similar phenotypic and genetic profiles. By leveraging these advanced methods, the platform supports personalized treatment strategies and the identification of drug combinations that maximize therapeutic potential.

[0046] FIG. 2 is a diagram illustrating an example of the gene description dataset module 102, in accordance with one or more embodiments of the current subject matter. The gene description dataset module 102 may include the gene identification module 202, data scraping engine 204, standardized query design module 206, and data cleaning and refinementDocket No.: 063697-505001WQ module 208 to work together in creating a comprehensive and high-quality gene description dataset.

[0047] The gene identification module 202 may be responsible for identifying a plurality of genes associated with a given disease. The gene identification module 202 may use data from various biological and genomic databases to compile a list of genes relevant to specific diseases. In some embodiments, the gene identification module 202 may identify a number of genes that are relevant enough to a disease, i.e., the genes that have a relevance level greater than a threshold score to a given disease. For example, the gene identification module 202 may identify the genes that exceeds the gene-disease-association (gDA) threshold scores in a literature or other scientific description. The gene identification module 202 may collaborate with the data scraping engine 204 to help identify the targeted genes and acquire associated descriptions. The data scraping engine 204 may function, for example, as an automated tool for extracting relevant gene descriptions from a variety of sources, such as scientific publications, genomic repositories, and medical databases. The data scraping engine 204 may work in coordination with the gene identification module 202 to focus on retrieving descriptions pertinent to the identified genes. The data scraping engine 204 may use web crawlers and API integrations to gather information efficiently. In some embodiments, the gene description dataset module 102 employs an automated data scraping engine 204 to extract relevant gene descriptions from various sources, including scientific publications, genomic databases, and clinical records. To maintain uniformity in the extracted data, a standardized query design module 206 formulates query prompts with specific parameters that define the context depth and length for each description. This ensures that the extracted data is comprehensive and consistent across different genes. The data cleaning and refinement module 208 processes the collected descriptions to remove any direct references to the specific disease in question, thereby preventing bias, data-leak, and maintaining the dataset's generalDocket No.: 063697-505001WQ applicability. This step involves natural language processing (NLP) techniques that detect and eliminate disease-specific mentions, ensuring that the resulting dataset supports unbiased downstream analyses and embedding processes.

[0048] The standardized query design module 206 may be tasked with designing and applying a consistent query prompt to extract descriptions for each identified gene. The standardized query design module 206 may help maintain the depth and context of the extracted data to be uniform across all genes. In some embodiments, the standardized query design module 206 may interface with natural language processing (NLP) tools to format and refine query prompts for optimal data extraction. In some embodiments, the prompt specifies consistent depth of context and length across the plurality of genes to maintain consistency across a number of descriptions for each gene associated with the disease. The data cleaning and refinement module 208 may process the extracted gene descriptions to remove any references to the specific disease in question. The data cleaning and refinement module 208 may maintain the dataset as generalized and unbiased by eliminating direct disease mentions, which may be useful for downstream analysis and embedding processes. The data cleaning and refinement module 208 may collaborate, for example, with the standardized query design module 206 to verify that only relevant content is retained while preserving the integrity of the extracted information. Removing references to the specific disease helps prevent potential bias in downstream applications, such as when embedding the data into high-dimensional vectors for comparative analysis. By omitting disease-specific references, the resulting dataset becomes more versatile and can be used for broader applications without introducing confounding factors linked to direct disease mentions that may result in tautological outputs. This approach supports a more objective evaluation of gene -drug relationships, enhancing the reliability of subsequent analyses and predictions within the platform. The gene identification module 202, data scraping engine 204, standardized query design module 206, and dataDocket No.: 063697-505001WG cleaning and refinement module 208, when integrated, may form a workflow that supports the generation of a gene description dataset, aiding in further processing and analysis within the platform.

[0049] FIG. 3 is a diagram illustrating an example of the text embedding engine 106, in accordance with one or more embodiments of the current subject matter. The text embedding engine 106 may include the preprocessing module 302, language model component 304, and output vector normalization module 306, which work in unison to transform textual data into high-dimensional vector representations for further analysis.

[0050] The preprocessing module 302 may prepare input data by performing standardization processes, such as tokenization, noise removal, and stop-word elimination. This process may help maintain the input data in a consistent format suitable for embedding. For example, the preprocessing module 302 may interface directly with the gene description dataset module 102 and MO A dataset module 104 to receive raw textual inputs and refine them for embedding. In some embodiments, algorithms for subgroup analysis, such as clustering algorithms (e.g., k-means, hierarchical clustering), are employed to analyze gene expression data correlated with distance scores. The platform 100 uses this analysis to identify specific patient subgroups that may respond favorably to particular drugs. For drug combination analysis, optimization algorithms evaluate potential drug pairs by calculating the combined coverage of target genes with favorable distance scores, ensuring broad therapeutic coverage. This approach enhances the identification of synergistic drug combinations that can address complex disease mechanisms more effectively.

[0051] The language model component 304 may utilize, in some embodiments, advanced language models, such as transformer-based architectures (e.g., BERT or GPT), to convert preprocessed gene descriptions and MOA data into vector representations. The language model component 304 may use models pre-trained on domain-specific biomedicalDocket No.: 063697-505001WG corpora to enhance the contextual representation of the vector outputs. The term "vector representations" refers to numerical formats that capture the semantic meaning of the input text, enabling quantitative analysis. This component may act as the core of the text embedding engine 106, providing the capability to represent complex biological text in a format suitable for analysis.

[0052] The output vector normalization module 306 may standardize the generated vectors to maintain consistency across the high-dimensional space. This process may involve scaling or adjusting the vector values so that they are comparable and ready for downstream distance calculations. The output vector normalization module 306 may, for example, support integration with the distance calculation module 108 by providing normalized vector representations. These components — preprocessing module 302, language model component 304, and output vector normalization module 306 — when combined, may enable the text embedding engine 106 to process and embed textual data efficiently, supporting the platform's approach for drug analysis and discovery processes.

[0053] FIG. 4 is a diagram illustrating a flow chart of a process 400 for ranking drugs for repurposing to treat a given disease, in accordance with one or more embodiments of the current subject matter. As shown in FIG. 4, the process 400 may begin with operation 402, where the system maintains a gene description dataset associated with a plurality of genes with respect to a given disease. This dataset may be generated using various data sources, including scientific publications and genomic databases, to compile comprehensive gene descriptions relevant to the disease.

[0054] In operation 404, the system maintains a dataset of drug mechanisms of action (MOAs) associated with a plurality of drugs. This MOA dataset may include detailed descriptions of how drugs interact with biological systems to exert their effects, collected and stored in a structured format to facilitate further analysis. The process may continue withDocket No.: 063697-505001WQ operation 406, where the system embeds the gene descriptions and the MO As into highdimensional vectors using a text embedding technique. This operation may leverage, in some embodiments, advanced language models such as BERT or GPT to generate vector representations that capture the semantic meaning of the input text, enabling gene and MOA descriptions to be quantitatively represented for comparative analysis. In operation 408, the system calculates distances between the gene description vectors and the MOA vectors using distance metrics such as cosine similarity, Euclidean distance, or Manhattan distance. This calculation may assess the relationships between genes and drug mechanisms, facilitating the evaluation of potential drug efficacy.

[0055] The process proceeds to operation 410, where the system ranks the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the disease. The ranking may be determined using an average or weighted average of the calculated distance scores to identify drugs that may be most relevant to treating the disease. In some embodiments, the process may include additional steps from dependent claims, such as generating the gene description dataset by identifying a plurality of genes associated with the given disease, scraping a plurality of sources to obtain gene descriptions, and applying a standardized query prompt to maintain consistent depth and context. The dataset may be refined by removing direct references to the disease to create an unbiased and generalizable dataset. The process may also involve refining the ranking by identifying drug combinations that maximize effectiveness across multiple genes or analyzing patient subgroups for whom specific drugs may be more effective. This analysis may include examining the variability of gene expression and distribution of distance scores to provide insights for personalized treatment approaches.

[0056] FIG. 5 depicts a block diagram illustrating a computing system 500 consistent with implementations of the current subject matter. As shown in FIG. 5, the computingDocket No.: 063697-505001WG system 500 can include a processor 510, a memory 520, a storage device 530, and input / output devices 540. The processor 510, the memory 520, the storage device 530, and the input / output devices 540 can be interconnected via a system bus 550. The computing system 500 may additionally or alternatively include a graphic processing unit (GPU), such as for image processing, and / or an associated memory for the GPU. The GPU and / or the associated memory for the GPU may be interconnected via the system bus 550 with the processor 510, the memory 520, the storage device 530, and the input / output devices 540. The memory associated with the GPU may store one or more images described herein, and the GPU may process one or more of the images described herein. The GPU may be coupled to and / or form a part of the processor 510. The processor 510 is capable of processing instructions for execution within the computing system 500. In some implementations of the current subject matter, the processor 510 can be a single-threaded processor. Alternately, the processor 510 can be a multi-threaded processor. The processor 510 is capable of processing instructions stored in the memory 520 and / or on the storage device 530 to display graphical information for a user interface provided via the input / output device 540.

[0057] The memory 520 is a computer-readable medium, such as volatile or nonvolatile memory, that stores information within the computing system 500. The memory 520 can store data structures representing configuration object databases, for example. The storage device 530 is capable of providing persistent storage for the computing system 500. The storage device 530 can be a floppy disk device, a hard disk device, an optical disk device, or a tape device, or other suitable persistent storage means. The input / output device 540 provides input / output operations for the computing system 500. In some implementations of the current subject matter, the input / output device 540 includes a keyboard and / or pointing device. In various implementations, the input / output device 540 includes a display unit for displaying graphical user interfaces.Docket No.: 063697-505001WQ

[0058] According to some implementations of the current subject matter, the input / output device 540 can provide input / output operations for a network device. For example, the input / output device 540 can include Ethernet ports or other networking ports to communicate with one or more wired and / or wireless networks (e.g., a local area network (LAN), a wide area network (WAN), the Internet).

[0059] In some implementations of the current subject matter, the computing system 500 can be used to execute various interactive computer software applications that can be used for organization, analysis and / or storage of data in various (e.g., tabular) format (e.g., Microsoft Excel®, and / or any other type of software). Alternatively, the computing system 500 can be used to execute any type of software applications. These applications can be used to perform various functionalities, e.g., planning functionalities (e.g., generating, managing, editing of spreadsheet documents, word processing documents, and / or any other objects, etc.), computing functionalities, communications functionalities, etc. The applications can include various add-in functionalities or can be standalone computing products and / or functionalities. Upon activation within the applications, the functionalities can be used to generate the user interface provided via the input / output device 540. The user interface can be generated and presented to a user by the computing system 500 (e.g., on a computer screen monitor, etc.).

[0060] One or more aspects or features of the subject matter described herein can be realized in digital electronic circuitry, integrated circuitry, specially designed framework specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) computer hardware, firmware, software, and / or combinations thereof. These various aspects or features can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and toDocket No.: 063697-505001WQ transmit data and instructions to, a storage system, at least one input device, and at least one output device. The programmable system or computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0061] These computer programs, which can also be referred to as programs, software, software frameworks, frameworks, components, or code, include machine instructions for a programmable processor, and can be implemented in a high-level procedural language, an object-oriented programming language, a functional programming language, a logical programming language, and / or in assembly / machine language. As used herein, the term “machine-readable medium” refers to any computer program product, apparatus and / or device, such as for example magnetic discs, optical disks, memory, and Programmable Logic Devices (PLDs), used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor. The machine-readable medium can store such machine instructions non-transitorily, such as for example as would a non-transient solid-state memory or a magnetic hard drive or any equivalent storage medium. The machine-readable medium can alternatively or additionally store such machine instructions in a transient manner, such as for example as would a processor cache or other random access memory associated with one or more physical processor cores.

[0062] To provide for interaction with a user, one or more aspects or features of the subject matter described herein can be implemented on a computer having a display device,Docket No.: 063697-505001WQ such as for example a cathode ray tube (CRT) or a liquid crystal display (LCD) or a light emitting diode (LED) monitor for displaying information to the user and a keyboard and a pointing device, such as for example a mouse or a trackball, by which the user may provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well. For example, feedback provided to the user can be any form of sensory feedback, such as for example visual feedback, auditory feedback, or tactile feedback; and input from the user may be received in any form, including, but not limited to, acoustic, speech, or tactile input. Other possible input devices include, but are not limited to, touch screens or other touch-sensitive devices such as single or multi-point resistive or capacitive trackpads, voice recognition hardware and software, optical scanners, optical pointers, digital image capture devices and associated interpretation software, and the like.Use case 1

[0063] In the search for alternative therapeutic applications, a team of researchers uses the platform to analyze drugs that did not pass FDA approval but have comprehensive preclinical and clinical data available. By leveraging the platform’s ability to maintain a dataset of drug mechanisms of action (MOAs) and a human-comprehensible gene description dataset generated by a Earge Eanguage Model (LLM), the team embeds these descriptions into high-dimensional vectors. The system then calculates distance scores between the gene description vectors and the MOA vectors using distance metrics such as cosine similarity or Euclidean distance to rank these drugs based on their aggregated distance scores. For example, the analysis identifies a cardiovascular drug, which had passed safety trials but failed efficacy tests for its original indication or failed to meet the clinical trial primary endpoint, as sharing strong mechanistic similarities with gene pathways implicated in a rare metabolic disorder. This insight is derived from ranking the drugs based on their proximity to gene vectors associated with the disorder, thus allowing the research team to prioritize drugsDocket No.: 063697-505001WQ with high potential for repurposing or refocusing. Equipped with this knowledge, the researchers can design targeted follow-up studies and implement a streamlined development process, potentially reviving the drug for a new therapeutic area with a higher likelihood of success.Use case 2

[0064] A research and development team working on a new immunotherapy drug for an autoimmune disease utilizes the platform to predict the likelihood of successful FDA approval. The platform maintains a comprehensive dataset of mechanisms of action (MO As) and a gene description dataset generated by a Large Language Model (LLM) to facilitate this analysis. By embedding both the MOA data and gene descriptions from clinical trial datasets into high-dimensional vectors using advanced text embedding techniques, the platform generates MOA vectors and gene description vectors. It then calculates distance scores between these vectors to compare the new drug’s characteristics with profiles of previously approved and failed drugs.

[0065] The analysis process leverages various distance metrics, such as cosine similarity and Euclidean distance, to quantify the alignment between the new drug’s MOA and historical data. The platform further aggregates these distance scores, calculating weighted averages to provide an overall assessment of the new drug’s potential alignment with successful cases. For instance, the system indicates that drugs with similar MOAs and gene interaction profiles demonstrated higher approval rates, especially when specific biomarkers were involved.

[0066] This insight allows the research team to refine their clinical trial design, ensuring it incorporates relevant biomarkers identified during the platform’s analysis. By aligning their trial parameters with characteristics found in previously successful drugs, theDocket No.: 063697-505001WQ team can optimize their regulatory strategy, increasing the probability of a favorable outcome. The ability to identify predictive markers and align the drug's profile with known success patterns exemplifies the platform’s strategic value in supporting data-driven decisionmaking for drug development and approval processes.Use case 3

[0067] A research team focused on developing combination therapies for oncology utilizes the platform to identify pairs of drugs that could potentially work synergistically for enhanced treatment efficacy. The platform leverages a comprehensive dataset of mechanisms of action (MO As) and gene expression data to facilitate this analysis. By embedding the gene expression data from patient tumor samples and MOA descriptions of various drugs into high-dimensional vectors using advanced text embedding techniques, the platform generates gene description vectors and MOA vectors for comparison.

[0068] The system calculates distance scores between these vectors using distance metrics such as cosine similarity and Euclidean distance to assess the alignment of combined drug mechanisms with the disease-specific genetic profile. The platform further aggregates these scores, computing averages or weighted averages to rank potential drug combinations based on their overall suitability for synergistic effects.

[0069] For example, the analysis reveals that pairing an experimental checkpoint inhibitor with a metabolic pathway modulator, each of which has demonstrated moderate efficacy as a single agent, could potentially amplify the immune response against cancer cells when combined. This insight provides the research team with a data-driven rationale to design targeted preclinical experiments and initiate pilot clinical trials to validate the effectiveness of the proposed combination therapy.Docket No.: 063697-505001WQ

[0070] The platform’s ability to analyze complex relationships between MO As and gene expression variability, and to identify high-potential drug combinations, exemplifies its value in accelerating the development of novel and more effective cancer treatment strategies. By leveraging patient-specific genetic data and assessing multi-drug interactions, the platform supports personalized and targeted approaches in oncology research and drug development.Use case 4

[0071] A university research team studying chronic inflammatory diseases employs the platform to identify shared pathophysiological mechanisms across multiple conditions. The platform, which integrates the use of a gene description dataset generated by a Large Language Model (LLM) and advanced text embedding techniques, processes these comprehensive gene descriptions related to various diseases. By embedding these descriptions into high-dimensional vectors, the platform organizes them into gene description vectors and clusters the vectors using machine learning algorithms such as hierarchical clustering and k-means clustering to generate a universal disease map.

[0072] The platform calculates distance scores between gene description vectors to assess the similarities and differences among pathophysiological features. This clustering analysis reveals common pathogenetic domains, such as “T-cell mediated” or “fibrotic” responses, as well as subdomains that span conditions like rheumatoid arthritis, systemic lupus erythematosus, and scleroderma. The identification of these shared mechanisms allows the research team to understand how different chronic inflammatory diseases may overlap in their underlying biological processes.

[0073] By highlighting these shared pathogenetic domains, the platform supports the researchers in identifying subtypes of diseases with similar mechanisms. This insight guidesDocket No.: 063697-505001WQ the team in exploring potential cross-disease treatments that target these shared pathways. For example, therapies that modulate T-cell responses or address fibrotic tissue formation could be applicable to multiple related diseases, providing a more unified treatment approach. The generated insights facilitate the development of cross-disease research strategies and help in designing broader therapeutic approaches that leverage common biological pathways. This capability of the platform to analyze complex relationships between gene expressions and disease mechanisms exemplifies its role in advancing personalized medicine and enhancing the scope of therapeutic research across different yet related inflammatory conditions.

[0074] In the descriptions above and in the claims, phrases such as “at least one of’ or “one or more of’ may occur followed by a conjunctive list of elements or features. The term “and / or” may also occur in a list of two or more elements or features. Unless otherwise implicitly or explicitly contradicted by the context in which it used, such a phrase is intended to mean any of the listed elements or features individually or any of the recited elements or features in combination with any of the other recited elements or features. For example, the phrases “at least one of A and B;” “one or more of A and B;” and “A and / or B” are each intended to mean “A alone, B alone, or A and B together.” A similar interpretation is also intended for lists including three or more items. For example, the phrases “at least one of A, B, and C;” “one or more of A, B, and C;” and “A, B, and / or C” are each intended to mean “A alone, B alone, C alone, A and B together, A and C together, B and C together, or A and B and C together.” Use of the term “based on,” above and in the claims is intended to mean, “based at least in part on,” such that an unrecited feature or element is also permissible.

[0075] The subject matter described herein can be embodied in systems, apparatus, methods, and / or articles depending on the desired configuration. The implementations set forth in the foregoing description do not represent all implementations consistent with the subject matter described herein. Instead, they are merely some examples consistent withDocket No.: 063697-505001WO aspects related to the described subject matter. Although a few variations have been described in detail above, other modifications or additions are possible. In particular, further features and / or variations can be provided in addition to those set forth herein. For example, the implementations described above can be directed to various combinations and subcombinations of the disclosed features and / or combinations and subcombinations of several further features disclosed above. In addition, the logic flows depicted in the accompanying figures and / or described herein do not necessarily require the particular order shown, or sequential order, to achieve desirable results. Other implementations may be within the scope of the following claims.

Claims

Docket No.: 063697-505001WOCLAIMSWhat is claimed is:

1. A computer-implemented method, the method comprising:maintaining a gene description dataset associated with a plurality of genes with respect to a given disease, wherein the gene description dataset is a human-comprehensible text-based dataset generated by a Large Language Model (LLM);maintaining a dataset of drug mechanisms of action (MO As) associated with a plurality of drugs;embedding the gene descriptions and the MO As into high-dimensional vectors using a text embedding technique to generate gene description vectors and MOA vectors;calculating distances between the gene description vectors and the MOA vectors to generate distance scores; andranking the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

2. The method of claim 1, wherein the gene description dataset is generated by:identifying a plurality of genes associated with the given disease;scraping a plurality of sources to obtain descriptions of each gene in relation to the given disease;designing and applying a standardized query prompt to extract descriptions for each gene associated with the given disease, wherein the prompt specifies consistent depth of context and length across the plurality of genes; andremoving references to the given disease from the extracted descriptions to create the gene description dataset.

3. The method of claim 1, wherein the text embedding technique comprises applying a language model to convert the gene descriptions and the drug MO As into vector representations in a high-dimensional space.

4. The method of claim 1, wherein calculating the distance between vectors comprises using one or more distance metrics selected from the group consisting of cosine similarity, Euclidean distance, Manhattan distance, and Minkowski distance.Docket No.: 063697-505001WO5. The method of claim 1, further comprising aggregating the distance scores by calculating an average or weighted average of the distances between the gene description vectors and MOA vectors to determine a suitability of each drug to the given disease.

6. The method of claim 1, wherein the ranking of drugs is further refined by identifying drug combinations that maximize a number of genes associated with the given disease with distance scores below a predetermined threshold.

7. The method of claim 1, further comprising identifying subgroups of patients for whom a particular drug is more effective by analyzing a distribution of distance scores and an expression variability of the genes associated with distinct phenotypes of the given disease.

8. A computer program product comprising a non-transient machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising:maintaining a gene description dataset associated with a plurality of genes with respect to a given disease, wherein the gene description dataset is a human-comprehensible text-based dataset generated by a Large Language Model (LLM);maintaining a dataset of drug mechanisms of action (MO As) associated with a plurality of drugs;embedding the gene descriptions and the MO As into high-dimensional vectors using a text embedding technique to generate gene description vectors and MOA vectors;calculating distances between the gene description vectors and the MOA vectors to generate distance scores; andranking the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

9. The computer program product of claim 8, wherein the operations further comprise:generating the gene description dataset by identifying a plurality of genes associated with the given disease;scraping a plurality of sources to obtain descriptions of each gene in relation to the given disease;Docket No.: 063697-505001WQdesigning and applying a standardized query prompt to extract descriptions for each gene associated with the given disease, wherein the prompt specifies consistent depth of context and length across the plurality of genes; andremoving references to the given disease from the extracted descriptions to create the gene description dataset.

10. The computer program product of claim 8, wherein the text embedding technique comprises applying a language model to convert the gene descriptions and the drug MO As into vector representations in a high-dimensional space.

11. The computer program product of claim 8, wherein calculating the distance between vectors comprises using one or more distance metrics selected from the group consisting of cosine similarity, Euclidean distance, Manhattan distance, and Minkowski distance.

12. The computer program product of claim 8, further comprising:aggregating the distance scores by calculating an average or weighted average of the distances between the gene description vectors and MOA vectors to determine a suitability of each drug to the given disease.

13. The computer program product of claim 8, wherein the operations further comprise:identifying drug combinations that maximize a number of genes associated with the given disease with distance scores below a predetermined threshold.

14. The computer program product of claim 8, wherein the operations further comprise:analyzing a distribution of distance scores and an expression variability of the genes associated with distinct phenotypes of the given disease to identify subgroups of patients for whom a particular drug is more effective.

15. A system comprising:at least one programmable processor; anda non-transient machine-readable medium storing instructions that, when executed by the processor, cause the at least one programmable processor to perform operations comprising:Docket No.: 063697-505001WOmaintaining a gene description dataset associated with a plurality of genes with respect to a given disease, wherein the gene description dataset is a human- comprehensible text-based dataset generated by a Large Language Model (LLM); maintaining a dataset of drug mechanisms of action (MO As) associated with a plurality of drugs;embedding the gene descriptions and the MOAs into high-dimensional vectors using a text embedding technique to generate gene description vectors and MOA vectors;calculating distances between the gene description vectors and the MOA vectors to generate distance scores; andranking the drugs for the given disease based on an aggregated distance score between the MOA vectors and the vectors of genes associated with the given disease.

16. The system of claim 15, wherein the operations further comprise:generating the gene description dataset by identifying a plurality of genes associated with the given disease;scraping a plurality of sources to obtain descriptions of each gene in relation to the given disease; designing and applying a standardized query prompt to extract descriptions for each gene associated with the given disease, wherein the prompt specifies consistent depth of context and length across the plurality of genes; and removing references to the given disease from the extracted descriptions to create the gene description dataset.

17. The system of claim 15, wherein the text embedding technique comprises applying a language model to convert the gene descriptions and the drug MOAs into vector representations in a high-dimensional space.

18. The system of claim 15, wherein calculating the distance between vectors comprises using one or more distance metrics selected from the group consisting of cosine similarity, Euclidean distance, Manhattan distance, and Minkowski distance.

19. The system of claim 15, wherein the operations further comprise:Docket No.: 063697-505001WQaggregating the distance scores by calculating an average or weighted average of the distances between the gene description vectors and MOA vectors to determine a suitability of each drug to the given disease.

20. The system of claim 15, wherein the operations further comprise:identifying drug combinations that maximize a number of genes associated with the given disease with distance scores below a predetermined threshold.

Images