Search

CN-121983125-A - System for predicting ICU patient outcome based on bronchoalveolar lavage fluid mNGS flora

CN121983125ACN 121983125 ACN121983125 ACN 121983125ACN-121983125-A

Abstract

The invention relates to the technical field of molecular biomedicine, in particular to a method and a system for predicting ICU patient outcome based on combination of bronchoalveolar lavage fluid mNGS flora and machine learning. The invention can provide a method for accurately predicting the ending of ICU patients for clinic.

Inventors

  • SHAO YANG
  • LIU JIA
  • ZHU MIAO
  • WU WEIWEI
  • Qi Baocui

Assignees

  • 迪飞医学科技(南京)有限公司
  • 南京迪飞医疗器械有限公司
  • 南京迪飞医学检验有限公司

Dates

Publication Date
20260505
Application Date
20251205

Claims (10)

  1. 1. A diagnostic system for predicting ICU patient outcome, comprising: A data processing module configured to receive metagenomic sequencing (mNGS) data of bronchoalveolar lavage fluid samples of a subject ICU patient and quantify abundance information of at least one strain marker selected from the following strain marker groups from the mNGS data to generate feature data; The strain marker group comprises streptococcus stomatitis (Streptococcus oralis), pasteurella (Pasteurellales), pasteurella (Pasteurellaceae), actinomycetes (Actinomyces), actinomycetes (Actinomycetaceae), actinomycetes (Actinomycetales), gingivalis (Eubacterium saphenum), european (Eubacterium), kidaceae (Mogibacteriaceae), streptococcus gastroenterology (Peptostreptococcaceae), pseudomonas revolve (Pseudomonas wittichii), lactococcus lactis (Lactococcus lactis), lactococcus (Lactobacillus), ralstonia (Ralstonia pickettii), acetobacter (Acetobacteraceae), rhodospirillum (Rhodospirillales), rosa mucosae (Roseomonas mucosa), rosa (Roseomonas) and Corynebacterium tuberculosis (Corynebacterium tuberculostearicum); the model prediction module is connected with the data processing module and is configured to input the characteristic data into a pre-constructed machine learning classification model, and calculate the predicted ending of the target ICU patient according to the model; and the result output module is connected with the model prediction module and is configured to output the predicted ending, wherein the predicted ending is survival or death.
  2. 2. The system of claim 1, wherein the data processing module is further configured to receive clinical signature data associated with the target ICU patient, and wherein the model prediction module is configured to combine the abundance information of the strain markers with the clinical signature data as a combined signature for input to the machine-learned classification model.
  3. 3. The system of claim 2, wherein the clinical profile data comprises at least one selected from the group consisting of number of hospitalization days, gender, age, HLA-DR, total lymphocyte count, cd3+ cells, cd3+cd4+ cells, cd3+cd8+ cells, B cells, NK cells, and cd4+/cd8+ ratio.
  4. 4. The system of claim 1, wherein the machine learning classification model is a random forest model, a gradient hoist model, a generalized linear model, a deep learning model, or a stacked model that integrates at least two of the foregoing models.
  5. 5. A computer-implemented method for predicting ICU patient outcome, comprising the steps of: a) Obtaining metagenomic sequencing (mNGS) data of a bronchoalveolar lavage fluid sample of a subject ICU patient; b) Quantifying abundance information of at least one species marker selected from the group of species markers defined in claim 1 from the mNGS data to obtain feature data; c) Inputting the characteristic data into a pre-constructed machine learning classification model, and calculating to obtain a predicted outcome of the target ICU patient; d) Outputting the predicted outcome, which is survival or death.
  6. 6. The method of claim 5, further comprising, after step b) and before step c): and acquiring clinical characteristic data of the target ICU patient, and combining the clinical characteristic data with the abundance information of the strain markers to jointly serve as characteristic data input into the machine learning classification model.
  7. 7. The method of claim 6, wherein the clinical profile data comprises at least one selected from the group consisting of number of hospitalization days, gender, age, HLA-DR, total lymphocyte count, cd3+ cells, cd3+cd4+ cells, cd3+cd8+ cells, B cells, NK cells, and cd4+/cd8+ ratio.
  8. 8. The method of claim 5, wherein the machine learning classification model is trained by a method comprising: i) Collecting bronchoalveolar lavage fluid samples from historical ICU patients and classifying them into a surviving group and a dead group according to their final outcome; ii) performing mNGS sequencing on the historical sample, and screening out the strain marker group through differential abundance analysis; iii) And training the machine learning classification model by using the abundance information of the strain markers as training features.
  9. 9. Use of a set of bacterial markers for the preparation of a system or kit for predicting ICU patient outcome, wherein the set of bacterial markers comprises streptococcus stomatis, pasteurella, pasteurellaceae, actinomycetes, gingivalis, euglena, kiwifruit, streptococcus peptic, revolve pseudomonas, lactococcus lactis, lactococcus, ralstonia, acetobacter, rhodospirales, rosacea, rhodomonas rosea, rhodomonas and corynebacterium tuberculosis; the use is to assess the risk of survival or death of a subject in an ICU patient to be tested by detecting the abundance of at least one marker in the set of bacterial markers in a bronchoalveolar lavage fluid sample.
  10. 10. The use according to claim 9, wherein the bacterial species marker is selected from at least one of Streptococcus stomatitis, pasteurella, actinomycetaceae, actinomycetales, guttiferae, euler's genus, leuconostoc family, and Streptococcus gastroenterology family, and wherein an increase in the abundance of the marker indicates an increased risk of the patient ending as survival; The bacterial species marker is selected from at least one of Pseudomonas revolve, lactococcus lactis, lactococcus, ralstonia, acetobacter, rhodospiriales, rosmarinomonas mucosae, rosomonas, and Corynebacterium tuberculosis, and an increase in the abundance of the marker indicates an increased risk of the patient ending as death.

Description

System for predicting ICU patient outcome based on bronchoalveolar lavage fluid mNGS flora Technical Field The invention relates to the technical field of molecular biomedicine, in particular to a method for accurately predicting ICU patient outcome based on mNGS flora and machine learning. Background Intensive Care Units (ICU) focus on the most critical patients in hospitals, whose disease progression is rapid and complex and variable, and whose survival outcome is comprehensively affected by a number of factors, including the severity of underlying disease, the status of organ function, the status of infection, and the effect of therapeutic intervention. Clinical predictions of ICU patient outcome (survival or death) have long been dependent primarily on traditional clinical indicators such as vital signs (heart rate, blood pressure, respiratory rate, etc.), laboratory tests (blood norms, liver and kidney function, inflammation index, etc.), and organ function scores (e.g., APACHE II scores, SOFA scores, etc.). However, these conventional methods have obvious limitations in that, on one hand, many clinical index changes often have hysteresis, and it is difficult to reflect prognosis trends in time in the early stages of critical turning of patient's condition, and on the other hand, single or partial index specificity is insufficient, and patients with different etiology and different stages may show similar index changes, resulting in limitation of prediction accuracy. With the development of molecular biology techniques, the metagenomic sequencing (mNGS) technique provides a powerful tool for the study of microbial communities. The method can directly sequence all microbial nucleic acids in the sample without culturing, comprehensively and quickly analyze the flora composition and pathogenic microorganism information in the sample, and has unique advantages in the aspects of pathogen diagnosis of infectious diseases and the like. Meanwhile, machine learning is used as an important branch of artificial intelligence, is increasingly widely applied to the aspects of disease prediction, diagnosis and the like in the medical field by virtue of strong data analysis and pattern recognition capability, and can mine potential rules and correlations from massive clinical and biological information. Although mNGS technology has advanced in both the field of microbiota analysis and machine learning in the field of medical prediction, currently mNGS microbiota analysis of bronchoalveolar lavage fluid (BALF) is combined with machine learning technology, and the related technology for ICU patient outcome prediction is still lacking. The flora state of the ICU patient lung is closely related to lung infection, immune response and the like, so that the prognosis of the patient can be influenced, flora characteristics in BALF are analyzed through mNGS technology, and a prediction model is constructed by combining machine learning, so that a new and more accurate method for ICU patient outcome prediction is hopefully provided, and the defects of the traditional prediction method are overcome. Disclosure of Invention According to the application, the microbial flora in the bronchoalveolar lavage fluid is analyzed, so that different species of different populations are screened out, a machine learning algorithm is adopted to incorporate clinical characteristics to construct a classifier, and the ending of an ICU patient is accurately predicted, so that the problems in the background technology are solved. A diagnostic system for predicting ICU patient outcome, comprising: A data processing module configured to receive metagenomic sequencing (mNGS) data of bronchoalveolar lavage fluid samples of a subject ICU patient and quantify abundance information of at least one strain marker selected from the following strain marker groups from the mNGS data to generate feature data; The strain marker group comprises streptococcus stomatitis (Streptococcus oralis), pasteurella (Pasteurellales), pasteurella (Pasteurellaceae), actinomycetes (Actinomyces), actinomycetes (Actinomycetaceae), actinomycetes (Actinomycetales), gingivalis (Eubacterium saphenum), european (Eubacterium), kidaceae (Mogibacteriaceae), streptococcus gastroenterology (Peptostreptococcaceae), pseudomonas revolve (Pseudomonas wittichii), lactococcus lactis (Lactococcus lactis), lactococcus (Lactobacillus), ralstonia (Ralstonia pickettii), acetobacter (Acetobacteraceae), rhodospirillum (Rhodospirillales), rosa mucosae (Roseomonas mucosa), rosa (Roseomonas) and Corynebacterium tuberculosis (Corynebacterium tuberculostearicum); the model prediction module is connected with the data processing module and is configured to input the characteristic data into a pre-constructed machine learning classification model, and calculate the predicted ending of the target ICU patient according to the model; and the result output module is connected with the model prediction module and