Fig. 6: Prediction and comparison of gene expression between S. lycopersicum and S. pennellii by the MSR model.
From: Deep learning the cis-regulatory code for gene expression in selected model plants
a Distribution of MSR model-predicted expression probabilities for the low (1st TPM quartile), middle (2nd and 3rd TPM quartiles) and high (4th TPM quartile) expressed transcripts of S. lycopersicum. Results of the shuffled-sequence control are plotted in grey for each respective expression level. b Distribution of MSR model-predicted expression probabilities for the low (1st TPM quartile), middle (2nd and 3rd TPM quartiles) and high (4th TPM quartile) expressed transcripts of S. pennellii. Results of the shuffled-sequence control are plotted in grey for each respective expression level. c Expression of MapMan functional categories exhibiting significant differential expression between S. lycopersicum and S. pennellii (Wilcoxon rank-sum test FDR <0.05). The difference between mean expression for the observed and predicted values provides the x and y axis coordinates, respectively. Data were scaled so that −1 indicates expression in S. lycopersicum only, 0 indicates equal expression in both species and 1 signifies unique expression in S. pennellii. Linear regression represents fit with a PCC of 0.56 (PCC p value 5.827e-09, R2 0.32). d Mapping of predicted expression probabilities (P; predicted) and observed gene expression (T; true labels) for enzymes of the polyamine biosynthesis pathway in S. lycopersicum and S. pennellii. Heatmaps are sorted according to gene orthologs, with the S. lycopersicum on the left side and the respective S. pennellii orthologs on the right.
