Researchers develop a new machine learning model to identify the type of cancer based on the tumors and the gene transformation.
The new artificial intelligence tool can identify the type of cancer with accuracy and higher degree of sensitivity. This newly developed deep-learning model identifies and may help to classify the cancers by analyzing the gene expression programs related to the early cell development and differentiation. However, parsing the differences in the gene expression is an ideal problem that requires a machine learning model.
However the machine learning model should be able to differentiate the healthy and cancer cells. If it is a complex model and has too much data on cancer gene expression, it may appear to learn the data perfectly. However, it would be malfunctioning when encountering a new set of data. Similarly, if the model is designed simply with a limited data set, it will lead to lack of information and does not serve the purpose.
The team focused on signs of altered developmental pathways in cancer cells to overcome this challenge. The reason is that the cancer cell will resemble the embryonic traits anytime and at any stage. The researchers compared the two large cell atlases – the Cancer Genome Atlas (TCGA) and the Mouse Organogenesis Cell Atlas (MOCA). The TCGA contains 33 tumor type gene expressions and the MOCA has 56 trajectories of embryonic cells. The model classed the tumors into four categories and yielded the information that could help with the diagnosis and treatment.
Though the study demonstrates a powerful approach to the tumor classification, it does have some limitations. The team of researchers is planning to increase the predictability of the model by incorporating more data from radiology, microscopy and other tumor imaging techniques.
