prescient-analysis

Notebooks for pre-processing, analysis, and visualization of PRESCIENT applied to in vitro hematopoeisis (Weinreb et al. 2020) and in vitro beta-cell differentiation (Veres et al. 2019).

• Current paper version: https://www.biorxiv.org/content/10.1101/2020.08.26.269332v1
• For in development PRESCIENT source code please visit https://github.com/gifford-lab/prescient/.

Organization

/scripts contains example bash scripts for fitting models:

• weinreb-interpolate.sh trains 5 seeds of 2 layer 400 unit models on only lineage tracing data with ground truth proliferation rates for the interpolation task on the Weinreb et al. dataset
• weinreb-fate.sh trains 5 seeds of 2 layer 400 unit models on all data with estimated proliferation rates for clonal fate bias prediction on the Weinreb et al. dataset
• veres-fate.sh trains 5 seeds of 2 layer 400 unit models with estimated proliferation rates on the Veres et al. dataset

/notebooks contains preprocessing, analysis and visualization workflows:

• 02{a-e}-weinreb2020-interpolation-* contains workflows for the interpolation task on the Weinreb et al. dataset
• 03{a-d}-weinreb2020-fate-* contains workflows for fate prediction task on the Weinreb et al. dataset
• 04{a-d}-weinreb2020-perturbations-* contains visualizations for perturbational experiments on the Weinreb et al. dataset
• 05{a-d}-veres2019-* contains workflows for fate prediction on the Veres et al. dataset
• 05{c-e}-veres2019-perturbations-* contains visualizations for timing perturbations on the Veres et al. dataset.
• 06{a-e}-veres2019-perturbations-* contains visualizations for large-scale perturbational screens and cell subset perturbations on the Veres et al. dataset.

Pre-processed data

Pre-processed data as generated by the workflows above can be downloaded from here

Pre-trained models

Pre-trained models used for perturbation experiments can be downloaded from here