Resumen
Datos solicitados: Datos brutos de RT-PCR realizadas para el diagnóstico de SARS-CoV-2
Objetivo del proyecto: Mejorar la fiabilidad de los diagnósticos por PCR
Requester
Name: Viktoria Lazar
Institution: Renyi Alfred Institute of Mathematics, Budapest, Hungary
Web: Machine learning PCR analysis
Description of project:
There is room for improvement in the evaluation of PCR tests. The raw output is a time series that has much more information than what we generally use. Instead of just checking whether the fluorescent signal crosses a certain threshold after a certain number of cycles, machine learning will help us to better exploit the information from the whole time series.
Terms of collaboration:
The requester is open to a long-term collaboration, including co-authorship of eventual papers. Besides providing the data, collaborators will have immediate access to the algorithm’s results, and may engage in testing and validating the platform that will be developed to make the algorithm easy to use by other diagnostic labs.
Detailed description of data:
We are working on a software in the cloud to analyze the RT-PCR diagnostic raw data using machine learning. In order to train our algorithm, we are looking for collaborators who are willing to provide us:
1) Raw data of their RT-PCR runs
2) The experts’ interpretation
Information about the data format can be found at open-pcr-analytics.org (please find the detailed information about the requested data files there). If manual work is involved to reformat the data, Crowdfight COVID-19 can provide volunteers to help with that task.
We would need ~100-150 positive samples and ~300-500 negative samples. The data can be anonymized, but it would be very useful to know whether the same person was tested more than once and the information about the date when the test was performed. We thought to collect this information because there are many examples when a given person is tested more than once and he/she turned to be positive only after the second or third test. So it is possible that we could use this information and identify the sample positive by our algorithm even in the first /second test when his/her sample contained a very low amount of virus particles.
Data can be transferred in one batch, or can be added slowly as new samples are processed (we are willing to engage in a long-term collaboration in which the data collection can be also prospective).
Protocol:
We will follow any requested policy. All this information will stay confidential and won’t be shared with a third part
Status of ethical approval:
The requester will work with anonymized datasets, and does not need ethical approval from their institution. They will work with the data provider to obtain any necessary ethical approvals to collect the data.
Other information:
Members of the project:
Endre Csoka (https://scholar.google.com/citations?user=6pNcwrYAAAAJ&hl=en)
Akos Torok (https://www.linkedin.com/in/%C3%A1kos-t%C3%B6r%C3%B6k-5a96377/)
Daniel Varga (https://scholar.google.com/citations?hl=en&user=x_5pH98AAAAJ)
Viktoria Lazar (https://scholar.google.com/citations?user=a7p4CvsAAAAJ&hl=en)
Tamas Nepusy (https://scholar.google.com/citations?user=wdyXgdEAAAAJ&hl=en)
Nota: En caso necesario, Crowdfight COVID-19 puede proporcionar ayuda para el procesamiento, preparación o envío de los datos. Esto incluye asesoramiento técnico, apoyo de expertos en bases de datos, o voluntarios que realicen tareas manuales (organización de datos, etc.).