### Motivation

See summary document here.

### Contents

- Big regression model
- Relevance of number of active bases in fragment
- CHiCAGO scores
- Improving the resolution of interactions
- GWAS Posterior Probabilities
- Emission state probabilities

To help decide which values of \(\tau\) to use in the QR analysis, I check the values at the quantiles of my data.

MPPC |
0.021 |
0.035 |
0.070 |
0.167 |

CHiCAGO |
1.610 |
2.533 |
5.270 |
15.875 |

PP |
0.002 |
0.010 |
0.051 |
0.407 |

### 1. Big regression

I run a multiple quantile regression using all of the emission proportions for a subset of 1,00,000 rows of the activated cell data.

The bait emission proportions all seem to have a small negative effect on MPPC, although most of these are not significant (especially for 95% quantile regression). Note that this sort of makes sense because these baited regions were *chosen* as they were near a TSS so are likely to be open anyway, whereas the prey fragment can be from anywhere in the genome.

The prey emission proportions all have varying effects, suggesting that the inference should be collapsed over several emission states.