Single Cell Analysis Results

[gwaybio]

in #77 we upgraded the single cell analysis to use data from batch 8 (data added in #75), and we use a multi-class classifier to predict clone A, clone E, and wildtype parental lines.

The results of this analysis were super interesting! Links to slides are below:

Preliminary analysis (#74) - https://docs.google.com/presentation/d/1_NwAVFJhqkA87duCBL5c2uxG9RlEj-qV4c5AFd4McWI/edit?usp=sharing
Batch 8 analysis (#77) - https://docs.google.com/presentation/d/13rOueMBlk8QbGx0G-Kl2qj42HVVShCbMF06PxXUfrv0/edit?usp=sharing

Primary Result

Many results and figures described in the batch 8 analysis slides relate to model performance and benchmarking model behavior. Below is the primary result to interpret.

Predicting wildtype parental single cells treated with various doses of bortezomib. Real data model compared to a model trained with shuffled data. As the drug dose increases, the clones appear more wildtype in nature, and thus, more likely to be killed by bortezomib treatment. The wildtype lines are not impacted by bortezomib treatment, indicating that the initial model is isolating the core differences between resistant and non-resistant lines, and that these features are not directly associated with bortezomib resistance.