Current COVID19 labs typically have the following procedure:

• 1. Acceptance of samples
• 2. Sample follow-up (LIMS etc.), logistics
• 3. qPCR testing
• 4. Data analysis with machine software and record the result in spreadsheet
• 5. Submit the results to some kind of government registry

Upon discussions, it seems like every step in this process craves for further optimization. With the genomize-covid19-qPCR software what we did targets the 4th step above. This step is error prone due to the 1) manual interpretation of curve structures of the qPCR results, where a sigmoid curve is expected, 2) manual recording of results in to spreadsheets. The sigmoid call for a curve is not a trivial issue, and especially for single channel kits this process may result to wrong test results. To addres this issue we used machine learning to train a classifier which considers several features of the curves deemed sigmoid, unsaturated sigmoid or non-sigmoid in the wet-laboratory tests performed at Turkish Health Institutions Federation (TÜSEB) COVID19 testing centers. We observed the following type of shapes:

• 1. Fully saturated sigmoid shaped curves, classified as sigmoid
• 2. The curves that has started saturation but not fully saturated, classified as sigmoid
• 3. The curves that has a twisted sigmoid shape, or early accelerating sigmoid curves as we call them, classified as sigmoid
• 4. The curves that did not start saturation, classified as unsaturated sigmoid, and tends to become negative in the retests.
• 5. Non-sigmoid curves, classified as non-sigmoid.

1,2 and 3 above were classified as sigmoid, 4 as unsaturated sigmoid, and 5 as non-sigmoid. This is a 3 class classification and trained with the data from the kits we worked with (Coyote COVID19 3 channel kit, and Sensure COVID19 3 channel kit) Below, you can see the graphics illustrating the above 5 classes. All feedbacks are wellcome at support@genomize.com. Important Note: We DON'T calculate CT values, we use machine software's calculations which is present in the file we require.

Different Sigmoid Shaped Curves

Unsaturated sigmoid shaped curves

Non-sigmoid sigmoid shaped curves