Genetic factors, risk prediction and AI application of thrombotic diseases

Abstract

In thrombotic diseases, coagulation, anticoagulation, and fibrinolysis are three key physiological processes that interact to maintain blood in an appropriate state within blood vessels. When these processes become imbalanced, such as excessive coagulation or reduced anticoagulant function, it can lead to the formation of blood clots. Genetic factors play a significant role in the onset of thrombotic diseases and exhibit regional and ethnic variations. The decision of whether to initiate prophylactic anticoagulant therapy is a matter that clinicians must carefully consider, leading to the development of various thrombotic risk assessment scales in clinical practice. Given the considerable heterogeneity in clinical diagnosis and treatment, researchers are exploring the application of artificial intelligence in medicine, including disease prediction, diagnosis, treatment, prevention, and patient management. This paper reviews the research progress on various genetic factors involved in thrombotic diseases, analyzes the advantages and disadvantages of commonly used thrombotic risk assessment scales and the characteristics of ideal scoring scales, and explores the application of artificial intelligence in the medical field, along with its future prospects.

Keywords: Artificial intelligence; Genetic factors; Machine learning; Risk prediction; Thrombophilia.

Fig. 1

The Virchow’s triad and the coagulant-anticoagulant-fibrinolytic system. The Virchow triad consists of vascular endothelial dysfunction, blood hypercoagulability and blood stasis. The risk factors of thrombotic diseases are divided into two categories: acquired and hereditary, as shown above. Hereditary factors mainly lead to hypercoagulable state. The balance of coagulation, anticoagulation and fibrinolysis makes the blood flow smoothly in the blood vessel and has the potential ability of thrombosis. Changes in the antigenic level or activity of any substance in this balance are likely to break the balance and lead to thrombosis

Fig. 2

Coagulation factor mutations lead to thrombosis. A Most of the mutations in prothrombin are single nucleotide substitutions (G20210A) at 20210 in the 3 ' non-coding region, resulting in abnormal 3 ' -terminal cleavage signals, accumulation of transcribed RNA, and corresponding increase in prothrombin produced at the translation level. B Prothrombin is a sodium-regulated allosteric enzyme. The sodium-binding domain consists of five amino acid residues: Thr540, Arg541, Glu592, Arg596 and Lys599.Single nucleotide substitution caused by mutations in this site easily leads to antithrombin resistance. C Arg506 is a kinetically favorable APC cleavage site, and the missense mutation at position 1691 causes Arg to be replaced by Gln (FV Leiden). On the one hand, FVa continues to express procoagulant activity, on the other hand, it resists the cleavage of APC, causing APC resistance-related thrombosis events. D The FIX R338L mutation was a gain-of-function mutation (FIX Padua), and the plasma level of FIX was normal, while the activity level was increased by about 8 times. At the same time, the stability of the combination with PS decreased, and the inhibition of PS decreased, showing a clear tendency of thrombosis

Fig. 3

Thrombosis caused by abnormal coagulation inhibitors. A 80% of the antithrombin deficiency is caused by SERPINC1 gene mutation. B In addition, N-glycosylation defects can lead to a decrease in its anti-FXa and anti-FIIa activity. C Tissue factor pathway inhibitor (TFPI) can specifically inhibit the tissue factor pathway and hinder the initiation of coagulation reaction. Single nucleotide substitution (C536T) in exon 7 will lead to a relative increase in the risk of venous thrombosis

Fig. 4

Thrombosis caused by abnormalities of fibrinolytic system and protein C system. A The carrying of 4G gene in the promoter region of PAI-1 coding gene increased the transcription level, protein expression and activity of PAI-1, and inhibited the dissolution of fibrin clot. B Mutations in protein C and protein S coding genes cause decreased activity and/or decreased antigen levels. Mutations in thrombomodulin will lead to a decrease in the affinity of the thrombin complex to protein C and a delay in the co-activation effect

Fig. 5

The basic process of using artificial intelligence to build medical models, and some common algorithms. The preprocessed data is usually divided into three parts: training set, validation set and test set. Researchers choose specific algorithms according to different application scenarios. After using the training set to learn and master the relevant knowledge, the corresponding models can be obtained. The validation set is used for verification, and the best model is selected after quantifying the score

Fig. 6

Application of AI in the medical field. The application fields of medical artificial intelligence include but are not limited to improving clinical diagnosis and treatment, assisting medical research, and promoting public health management. Application of clinical diagnosis and treatment: risk assessment, auxiliary diagnosis, clinical management, improvement of patient consultation, determination of surgical indications, intraoperative decision-making, postoperative management, etc. Auxiliary medical research: disease target prediction, drug component screening, discovery of disease rules, etc. Public health management: Internet hospitals, epidemic monitoring, optimization of medical resource allocation

Fig. 7

Combination of artificial intelligence and imaging. The combination of artificial intelligence and imaging can help to select examination methods, realize automatic report interpretation, individualized diagnosis, individualized treatment, prognosis evaluation and optimize health management. A case study of artificial intelligence to establish an automatic impact reporting system, which uses a variety of artificial intelligence technologies such as natural language processing and computer vision systems