Development of a software platform for studying the protein expression of microarrays

Introduction

Microarrays can house thousands of assays. The multiplexed output from each ligand is used to study the protein expression for a particular disease. The intensity extraction exercise, however, has to be performed manually for each spot. This not only delays the experiment, but is also prone to human errors. This work presents an automated technique to segregate the spot intensities from the background that has been embedded in an application software.

Methodology

The user interface also enables the operator to attach a file containing ligand position details. This generates an output file containing the protein names along with their respective intensities and other statistical parameters. The "Project" mode allows the user to load multiple images pertaining to the same assay and label them in subgroups as per the cohort or severity. In this mode, significant proteins or peptides are found and required visualizations are generated. User is given an option to set various parameters in the tasks of intensity extraction, statistical analysis and visualization.

The imaging process involves the denoising of the image to remove the features containing abrupt intensity changes. Then the image is segmented into object and background using the local threshold instead of global. Out of the extracted objects, microarray spots are selected based on the criteria of convolution. The row and column details of the spots are identified to match them with the provided file.

Statistical process begins with the preprocessing steps like background correction followed by the normalization or standardization techniques. The processed data is subjected to parametric or non-parametric according to the user preference to identify the novel biomarkers.

Conclusion

The said technique has not only been tried on protein and peptide microarrays scanned under high-end instrument but it can also extract data from microarray chips imaged under indigenous instrument. Unlike the contemporary methods, the feature pixels are not enclosed inside regular grid cells like circle or rectangle but their boundary contours are extracted ignoring the background area.

References

1. Ray, Arka, Sanjeeva Srivastava, and Siddhartha P. Duttagupta. "MicroNET: Microarray Image Orientation Predictor and Automated Data Extraction." In 2024 IEEE 3rd International Conference on Control, Instrumentation, Energy & Communication (CIEC), pp. 118-123. IEEE, 2024.

2. Acharjee, Arup, Arka Ray, Akanksha Salkar, Surbhi Bihani, Chaitanya Tuckley, Jayanthi Shastri, Sachee Agrawal, Siddhartha Duttagupta, and Sanjeeva Srivastava. "Humoral immune response profile of COVID-19 reveals severity and variant-specific epitopes: Lessons from SARS-CoV-2 peptide microarray." Viruses 15, no. 1 (2023): 248.