Search

CN-121983308-A - SLC31A 1-based castration-resistant prostate cancer prognosis prediction model

CN121983308ACN 121983308 ACN121983308 ACN 121983308ACN-121983308-A

Abstract

The invention discloses a castration resistance prostate cancer prognosis prediction model based on SLC31A1, and belongs to the technical field of bioinformatics and tumor accurate medical treatment. The model is realized through the following data processing flow, namely, molecular and clinical data of a patient at cross time points are obtained, functional activity integrated values at each time point are calculated and a time sequence curve is constructed by fusing transcription expression quantity of SLC31A1 and methylation level of a specific regulation region of the SLC, dynamic morphological characteristics of the curve are extracted and are subjected to matching analysis with a poor prognosis mode characteristic library pre-stored on the basis of historical data, the obtained matching degree, the current activity integrated value and the clinical characteristics are input into a decision function, quantized disease progression risk level is output, and trend suggestions of a treatment strategy can be further provided. The invention realizes the dynamic, quantitative and interpretable accurate prediction of the disease progression risk of the castration resistant prostate cancer patient, and provides basis for individualized treatment decision.

Inventors

  • HE JIAWEI
  • LI PEIZHEN

Assignees

  • 何嘉炜

Dates

Publication Date
20260505
Application Date
20260119

Claims (10)

  1. 1. A model for prognosis of castration-resistant prostate cancer based on SLC31A1, wherein the internal data processing logic comprises the steps of: S1, inputting cross-time-point molecules and clinical data of a target patient; S2, calculating an SLC31A1 functional activity integration value at each time point based on the molecular data, wherein the calculation of the value fuses the SLC31A1 mRNA expression quantity at the corresponding time point and the methylation level of a specific CpG site positioned in the promoter region of the SLC31A1 gene, and the methylation level participates in fusion by negative weight; S3, constructing the SLC31A1 functional activity integrated value at each time point into an activity time sequence curve according to time sequence, and extracting dynamic morphological characteristics of the curve, wherein the dynamic morphological characteristics comprise key turning points of the curve, slope values of each curve segment and accumulated deviation values of the curve relative to a preset base line; S4, performing matching degree calculation on the extracted dynamic morphological characteristics and a pre-stored bad prognosis mode characteristic library, wherein the bad prognosis mode characteristic library comprises a morphological characteristic set extracted from an activity time sequence curve of a bad patient in a history queue; S5, outputting a quantized prognosis risk level for the target patient through a predefined risk stratification rule based on the matching degree and the SLC31A1 functional activity integrated value of the current time point.
  2. 2. The model of claim 1, wherein in step S2, the calculation of the integrated value of the functional activity of SLC31A1 further introduces a third variable, which is the expression level of a specific microribonucleic acid in known regulatory relationship with SLC31 A1.
  3. 3. The SLC31 A1-based castration resistant prostate cancer prognosis model of claim 1, wherein in step S3, the extracting dynamic morphological features specifically includes: s3.1, identifying a point with zero first derivative or a point with extreme value of second derivative in the active time sequence curve as the key turning point; S3.2, calculating the average slope of a curve segment defined by adjacent key turning points as the segmentation slope value; And S3.3, calculating the area of a region surrounded by the active time sequence curve and a straight line connecting the starting point and the end point of the curve, and taking the area as the accumulated deviation value.
  4. 4. The SLC31 A1-based castration resistant prostate cancer prognosis prediction model of claim 1, wherein the pre-stored pool of poor prognosis pattern features in step S4 comprises: The first mode is characterized in that the slope value of the active time sequence curve between two continuous detection time points is larger than a first preset slope threshold value; the second mode is characterized in that at least three monotonically increasing steps exist in the activity time sequence curve, and the serum prostate specific antigen level recorded at the first clinical detection time point after each step is increased by more than a preset proportion threshold value compared with the recorded value at the previous clinical detection time point.
  5. 5. The model of SLC31 A1-based castration resistant prostate cancer prognosis of claim 4, wherein the matching calculation in step S4 is a weighted calculation in which the dynamic morphological features extracted during the latter 50% of the time span of the activity timing curve are weighted higher in calculation than features extracted during the first 50% of the time span.
  6. 6. The model of SLC31 A1-based castration resistant prostate cancer prognosis of claim 1, wherein the predefined risk stratification rule in step S5 is a decision function whose input variables include the degree of match, the SLC31A1 functional activity integration value at the current time point, and the gleason score in the clinical data.
  7. 7. The model of SLC31 A1-based castration resistant prostate cancer prognosis of claim 1, further comprising step S6: s6, if the quantized prognosis risk level output in the step S5 is a preset high risk level, executing a treatment strategy tendency deducing step, and outputting a corresponding priority recommended treatment strategy category based on the specific poor prognosis mode matched with the dynamic morphological characteristics.
  8. 8. The model for predicting prognosis of castration resistant prostate cancer based on SLC31A1 of claim 7, wherein the treatment strategy propensity inference step in step S6 is based on the association rule: If the dynamic morphological characteristics match the first mode characteristics, outputting the priority recommended treatment strategy category as copper metabolism intervention treatment; And if the dynamic morphological characteristics are matched with the second mode characteristics, outputting the priority recommended treatment strategy category which is deoxyribonucleic acid damage repair targeted treatment.
  9. 9. The model of claim 1, wherein the specific CpG site in the SLC31A1 gene promoter region in step S2 is one CpG dinucleotide site located within a region of 200 base pairs to 50 base pairs upstream of the SLC31A1 gene transcription initiation site.
  10. 10. The model of claim 1, wherein the negative weighting in step S2, the poor prognosis pattern feature library and matching threshold in step S4, and the risk stratification rule parameter in step S5 are all trained by a machine learning model from a time series training dataset of castration resistant prostate cancer patients comprising known disease progression time and total survival data.

Description

SLC31A 1-based castration-resistant prostate cancer prognosis prediction model Technical Field The invention relates to the technical field of bioinformatics and accurate tumor medical treatment, in particular to a castration resistance prostate cancer prognosis prediction model based on SLC31A 1. Background Tumor prognosis is an important basis for clinical decisions. Currently, biomarker-based prognostic assays mostly employ a static analysis model, i.e., detecting the expression level of a particular gene at a single point in time (usually at the time of diagnosis), and constructing a predictive model in conjunction with clinical pathological parameters. Such methods have the limitations that firstly, the dynamic evolution process of the molecular characteristics of the tumor in the disease process, especially under therapeutic intervention, is not reflected, and the process is directly related to the therapeutic response and disease progression, and secondly, the method generally depends on single-dimension information (such as the abundance of mRNA expression) and fails to integrate other key regulatory information (such as the methylation modification of deoxyribonucleic acid) affecting the functional state of the gene, so that the stability and universality of the marker can be affected. In the prior art, chinese patent CN116859048A discloses the application of transmembrane protein SLC31A1 as a tumor prognosis marker, which indicates that the expression level of the protein is related to prognosis of various tumors including prostate cancer. This patent points out the prognostic value of SLC31A 1. However, the technical solution still falls under the above-mentioned static analysis paradigm, and the claims thereof mainly relate to detecting the expression level (protein or nucleic acid) of SLC31A1 at a single point in time. This protocol did not involve analysis of epigenetic regulatory aspects of SLC31A1 function (such as promoter methylation), nor did it utilize timing detection data of the same patient at multiple treatment or follow-up nodes. Thus, in the face of a disease with highly heterogeneous and dynamically evolving characteristics, castration-resistant prostate cancer, the prior art solutions have room for further improvement in terms of accuracy, stability of predictions and insight in providing a temporal level. Disclosure of Invention In order to overcome the defects of the prior art, the embodiment of the invention provides a castration resistance prostate cancer prognosis prediction model based on SLC31A1 to solve the problems in the prior art, in particular to a scheme that the prior art is difficult to integrate multidimensional regulation information of SLC31A1 and analyze the rule of dynamic change of the regulation information along with time, and is difficult to output a risk assessment result which is more relevant to clinical decision. In order to achieve the aim, the invention provides the technical scheme that the castration resistant prostate cancer prognosis prediction model based on SLC31A1 comprises the following steps of: S1, inputting cross-time-point molecules and clinical data of a target patient; S2, calculating an integrated value of the SLC31A1 functional activity at each time point based on molecular data, wherein the calculation of the integrated value fuses the SLC31A1 mRNA expression quantity at the corresponding time point and the methylation level of a specific CpG site positioned in a promoter region of the SLC31A1 gene, and the methylation level participates in fusion by negative weight so as to comprehensively characterize the SLC31A1 functional activity regulated by epigenetic control; S3, constructing the SLC31A1 functional activity integrated value at each time point into an activity time sequence curve according to time sequence, and extracting dynamic morphological characteristics of the curve, wherein the dynamic morphological characteristics comprise key turning points of the curve, slope values of each curve section and accumulated deviation values of the curve relative to a preset base line so as to quantify a dynamic mode of activity change; S4, performing matching degree calculation on the extracted dynamic morphological characteristics and a pre-stored bad prognosis mode characteristic library, wherein the bad prognosis mode characteristic library comprises morphological characteristic sets extracted from activity time sequence curves of bad patients in a history queue, and identifying whether the current patient presents a known bad prognosis evolution track or not through matching; S5, outputting a quantized prognosis risk level for the target patient through a predefined risk stratification rule based on the matching degree and the SLC31A1 functional activity integrated value of the current time point. Further, in step S2, a third variable is introduced into the calculation of the integrated value of the functional activity of the SLC31A1