Simplified model fit to construct UTR data?

[ewallace]

For the chimeric UTR data analysis, all our constructs have the same 3'UTRs, and there are few motifs present, so we should fit a model to only that data.

At the moment the analysis in raw_data_analysis/pRPS3_pPGK1_tRPS3_two_exp_rep_RNA_abundance_with_model_pred.Rmd seems too complicated, and we should not need to download the coding sequences of yeast ORFs or to run the whole-transcriptome model fit in order to analyse this data for this figure.

There are two ways that we could simplify this:

1. Export the coefficients for codons and motifs from the whole-transcriptome fit. Then import them to this script to use for prediction based on the motif content.
   • This could predict the half-life of the constructs based on the codons and motifs
   • But we measure relative abundance, so we only need to estimate delta abundance using the subsets of motifs in the constructs.
2. Fit a simplified model with these motifs directly to the data. I.E. transform abundance data to log2-half-life-scale, then fit with glm.
   • then we could compare the coefficients obtained from that with the coefficients obtained from the whole-transcriptome model

[DimmestP]

As of 7th April I have not changed the the model code. I decided to prioritise finishing writing the manuscript, then return to simplifying the code.