Severe neutropenia caused by venetoclax,a B-cell lymphoma-2(BCL-2) inhibitor, is the main cause of venetoclax tapering, drug discontinuation, and treatment delay. This study combines machine learning and genomics, hoping to develop models to predict venetoclax dose in Acute myeloid leukemia(AML) patients and compare the efficacy and safety differences of model-guided individualized medication regimen with current conventional regimen. According to the demographic information, the drug information, the drug concentration of the target patients, the laboratory examination, the single nucleotide polymorphism(SNP) information and the adverse reactions of the AML patients, and the model was constructed through machine learning.

Introduction:The successful development of venetoclax offers new hope for AML patients not eligible for strong induction chemotherapy. However, there are some clinical problems, such as severe neutropenia is the main reason for treatment delay and discontinuation of patients. The Asian population has higher drug exposure than the non-Asian population, and the blood concentration of venetoclax varies greatly individually, and the blood drug concentration is associated with efficacy and adverse effects. We urgently need an individualized study of venetoclax for Chinese AML patients to reduce the incidence of adverse events while ensuring efficacy.

Objective：Construction of a venetoclax dose prediction model for AML patients based on machine learning combined genomics;

Methods：1.Venetoclax plasma concentration determination;determination of SNPs of related genes in patient blood cells; 2.venetoclax dose prediction model for AML patients based on machine learning techniques combined with genomics Collect the clinical data and establish a database Mining variables to explore the factors affecting the dosage of venetoclax Building a predictive model based on a machine-learning algorithm Model performance was evaluated, and the optimal model was selected Interpretation and optimization of the model

The AML patients were conditionally screened by the study physician involved in the project department to assess their enrollment. Communicate fully with the patients and their family members who meet the enrollment criteria, obtain the patient's informed consent, and sign the informed consent form. After enrollment, patient clinical data were recorded. Evaluation according to the efficacy and safety evaluation criteria.

Inclusion Criteria:

1. Age ≥ 18 years old, regardless of gender;
2. Diagnosed as an AML patient according to the Diagnosis and Treatment Guidelines for Adult Acute Myeloid Leukemia (Non Acute Promyelocytic Leukemia) in China (2021 Edition) and receiving treatment with venetoclax;
3. Before receiving venetoclax treatment, absolute neutrophil count (ANC) ≥ 1.0 ×10 ^9/L, white blood cell count (WBC) ≥ 2.0 ×10 ^9/L, platelet count (PLT) ≥ 50 ×10 ^9/L, and hemoglobin (HB) ≥ 90g /L;
4. Before receiving venetoclax treatment, liver and kidney function were normal (aspartate aminotransferase ≤ 3 times the upper limit of normal (ULN), alanine aminotransferase ≤ 3.0 x ULN, bilirubin ≤ 1.5 x ULN, urea nitrogen:3.2-7.1 mmol/L, glomerular filtration rate (eGFR) ≥ 60ml/min;
5. Sign an informed consent form.

Exclusion Criteria:

1. Age<18 years old;
2. Non AML patients;
3. Patients who plan to use a treatment regimen without venetoclax;
4. Patients with poor medication adherence;
5. Liver and kidney function damage before medication;
6. Before medication, ANC<1.0 x 10 ^9/L or WBC<2.0 x 10 ^9/L or PLT<50 x 10 ^9/L or HB<90g /L;
7. Pregnant and lactating women;
8. Cases deemed unsuitable for inclusion by researchers

This study aims to include 200 patients with AML who were treated with venetoclax. A database of venetoclax medication for AML patients will be constructed based on demographic information, medication information, laboratory test information, blood drug concentration, SNPs, and drug-related adverse reactions. Randomly divide the data into training and testing sets in a 7:3 ratio, and construct the model through machine learning.