| CPC G16H 50/20 (2018.01) [G01N 33/5011 (2013.01); G16B 5/20 (2019.02); G16B 40/20 (2019.02)] | 10 Claims |

|
1. A method for training a predictive model the method comprising:
a) obtaining an initial datasets from at least one of wild type tumor cell lines, a genetically engineered tumor, a patient dataset associated with a specific tumor type from at least one of wild type tumor cell lines and genetically engineered tumor;
b) normalizing the initial dataset in accordance with the respective patient dataset;
c) generating first correlations between the initial dataset and a first output indicative of whether the tumor is a metastatic tumor, by assigning weightages to each parameter of the initial dataset using classified supervised learning and selecting an optimized set of parameters from datasets comprising-chemosensitivity, stemness, adhesion, exosome-uptake, autophagy, migration, angiogenesis, and spheroid formation;
d) mapping the first correlations by associating the initial dataset and the patient dataset with historical training data to refine the predictive model;
e) dividing the initial dataset and the patient dataset into a training set, a validation set, and a test set; and
f) iteratively training, validating, and testing, the predictive model using the training set; and
g) dynamically adjusting the first correlations based on the assigned weightages using the predictive model.
|