Subio Platform working with ChatGPT and Jupyter.
Subio Platform can import and export files in tab-delimited text format. This allows you to work with other tools to complement functions that are not included in Subio Platform or its plug-ins. If you can write R scripts, there is no problem. You can combine the interface and data management functions of Subio Platform with the statistical analysis functions of Bioconductor .
On the other hand, even if you cannot code, recent advances in generative AI have made it possible to ask ChatGPT to generate Python code and run it in Jupyter. This video shows how to do that.
The key to having ChatGPT write useful code is to clearly describe the specifications and algorithm without leaving out important details. ChatGPT may ask you to upload a data file. In such cases, ask it to generate Python code instead.
Jupyter provides a simple user interface and allows you to run Python code without setting up a full development environment. You can download the generated text file and import it into Subio Platform for use in your analysis.
Overview of the Procedure Shown in This Video
This video introduces a simple example of using a script generated with ChatGPT to classify samples in a way that is not easy to perform directly within Subio Platform alone.
In this example, we use two gene lists: base genes and target genes. For each sample, we first calculate the median expression value of the base genes. Then, among the genes included in the target genes list, we count how many genes show expression values higher than that median.
Based on this count, each sample is classified using a simple rule. By importing the classification results back into Subio Platform as experimental parameters, you can assign new labels to the samples.
In this way, external scripts such as R or Python can be used not only for advanced statistical analysis, but also for sample classification and data processing based on simple decision rules. By bringing the results back into Subio Platform, you can use them for downstream visualization and comparison, such as PCA, clustering, heat maps, and group comparisons.
The important point is that you do not need to build a large analysis pipeline from the beginning. By looking at your data in Subio Platform and adding only the necessary processing as small external scripts, you can extend the flexibility of your analysis.