1. How can I interpret the results?
  2. What types of cells were used to train DNN in Embryonic.ai?
  3. What gene expression platforms are compatible with Embryonic.ai?
  4. Can I use RNA-Seq data in Embryonic.ai?
  5. Can I try my mouse embryonic stem cell data?
  6. How was Embryonic.ai built?
  7. How was the training data set generated?

Q: How can I interpret the results?
This long green progressbar is showing the embryonic score (E score) of the sample.
E score acts as an integrative measure of cell development stage, where ES=1 and ES=0 is representative of a embryonic and adult cellular states, respectively.

Q: What types of cells were used to train DNN in Embryonic.ai?
A: We used healthy

Q: What expression platforms are compatible with Embryonic.ai?

A: Embryonic.ai was designed for working with both Illumina and Affymetrix platforms. However it can be used in other platform types. For the most popular platforms Embryonic.ai performs automatic preprocessing, so the user needs to specify only GSM identifier. Samples from following popular platforms (including alternative names) are preprocessed automatically:

Q: Can I use RNA-Seq data in Embryonic.ai?
A: Yes, you can upload RNA-Seq data as a custom file with Upload file option. From our experience, when working with RNA-Seq data it is best to choose Affymetrix DNN for the analysis.

Q: Can I try my mouse embryonic stem cell data?

A: Yes, it is possible to use the mouse gene expression data, where mouse orthologs’ gene symbols are converted to human HGNC gene symbol format. But it must be considered that the system was trained on human data, so the results should be examined with caution.

Q: How was Embrionic.ai built?
A: Training gene expression data was collected and preprocessed independently for Affymetrix and Illumina platforms. Then a set of DNN classifiers were trained and the ones showing best performance were chosen and put in an ensemble.

Q: How was the complete training set generated?
A: We used data from two public databases: Gene Expression Omnibus (GEO) and ArrayExpress, as well as additional dataset provided by BioTime, Inc.