High-resolution temporal profiling of E. coli transcriptional response

Understanding how cells dynamically adapt to their environment is a primary focus of biology research. Temporal information about cellular behavior is often limited by both small numbers of data time-points and the methods used to analyze this data. Here, we apply unsupervised machine learning to a data set containing the activity of 1805 native promoters in E. coli measured every 10 minutes in a high-throughput microfluidic device via fluorescence time-lapse microscopy. Specifically, this data set reveals E. coli transcriptome dynamics when exposed to different heavy metal ions. We use a bioinformatics pipeline based on Independent Component Analysis (ICA) to generate insights and hypotheses from this data. We discovered three primary, time-dependent stages of promoter activation to heavy metal stress (fast, intermediate, and steady). Furthermore, we uncovered a global strategy E. coli uses to reallocate resources from stress-related promoters to growth-related promoters following exposure to heavy metal stress.


Fig
Fig S1|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 0. Bottom: Genes weights above threshold for IModulon 0.

Fig
Fig S2|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 1. Bottom: Genes weights above threshold for IModulon 1.

Fig
Fig S3|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 2. Bottom: Genes weights above threshold for IModulon 2.

Fig
Fig S4|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 3. Bottom: Genes weights above threshold for IModulon 3.

Fig
Fig S5|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 4. Bottom: Genes weights above threshold for IModulon 4.

Fig
Fig S6|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 5. Bottom: Genes weights above threshold for IModulon 5.

Fig
Fig S7|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 6. Bottom: Genes weights above threshold for IModulon 6.

Fig
Fig S8|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 7. Bottom: Genes weights above threshold for IModulon 7.

Fig
Fig S9|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 8. Bottom: Genes weights above threshold for IModulon 8.

Fig
Fig S10|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 9. Bottom: Genes weights above threshold for IModulon 9.

Fig
Fig S11|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 10.Bottom: Genes weights above threshold for IModulon 10.

Fig
Fig S12|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 11.Bottom: Genes weights above threshold for IModulon 11.

Fig
Fig S13|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 12. Bottom: Genes weights above threshold for IModulon 12.

Fig
Fig S14|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 13.Bottom: Genes weights above threshold for IModulon 13.

Fig
Fig S15|ICA results from running the algorithm on the induction window time points.Top: Activation profile plot of iModulon 14.Bottom: Genes weights above threshold for IModulon 14.

Fig
Fig S16|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 0. Bottom: Genes weights above threshold for IModulon 0.

Fig
Fig S17|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 1. Bottom: Genes weights above threshold for IModulon 1.

Fig
Fig S18|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 2. Bottom: Genes weights above threshold for IModulon 2.

Fig
Fig S19|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 3. Bottom: Genes weights above threshold for IModulon 3.

Fig
Fig S20|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 4. Bottom: Genes weights above threshold for IModulon 4.

Fig
Fig S21|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 5. Bottom: Genes weights above threshold for IModulon 5.

Fig
Fig S22|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 6. Bottom: Genes weights above threshold for IModulon 6.

Fig
Fig S23|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 7. Bottom: Genes weights above threshold for IModulon 7.

Fig
Fig S24|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 8. Bottom: Genes weights above threshold for IModulon 8.

Fig
Fig S25|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 9. Bottom: Genes weights above threshold for IModulon 9.

Fig
Fig S26|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 10.Bottom: Genes weights above threshold for IModulon 10.

Fig
Fig S27|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 11.Bottom: Genes weights above threshold for IModulon 11.

Fig
Fig S28|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 12. Bottom: Genes weights above threshold for IModulon 12.

Fig
Fig S29|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 13.Bottom: Genes weights above threshold for IModulon 13.

Fig
Fig S30|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 14.Bottom: Genes weights above threshold for IModulon 14.

Fig
Fig S31|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 15.Bottom: Genes weights above threshold for IModulon 15.

Fig
Fig S32|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 16.Bottom: Genes weights above threshold for IModulon 16.

Fig
Fig S33|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 17.Bottom: Genes weights above threshold for IModulon 17.

Fig
Fig S34|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 18. Bottom: Genes weights above threshold for IModulon 18.

Fig
Fig S35|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 19.Bottom: Genes weights above threshold for IModulon 19.

Fig
Fig S36|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 20.Bottom: Genes weights above threshold for IModulon 20.

Fig
Fig S37|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 21.Bottom: Genes weights above threshold for IModulon 21.

Fig
Fig S38|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 22. Bottom: Genes weights above threshold for IModulon 22.

Fig
Fig S39|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 23.Bottom: Genes weights above threshold for IModulon 23.

Fig
Fig S40|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 24.Bottom: Genes weights above threshold for IModulon 24.

Fig
Fig S41|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 25.Bottom: Genes weights above threshold for IModulon 25.

Fig
Fig S42|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 26.Bottom: Genes weights above threshold for IModulon 26.

Fig
Fig S43|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 27.Bottom: Genes weights above threshold for IModulon 27.

Fig
Fig S44|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 28.Bottom: Genes weights above threshold for IModulon 28.

Fig
Fig S45|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 29.Bottom: Genes weights above threshold for IModulon 29.

Fig
Fig S46|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 30.Bottom: Genes weights above threshold for IModulon 30.

Fig
Fig S47|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 31.Bottom: Genes weights above threshold for IModulon 31.

Fig
Fig S48|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 32.Bottom: Genes weights above threshold for IModulon 32.

Fig
Fig S49|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 33.Bottom: Genes weights above threshold for IModulon 33.

FigFig
Fig S50|ICA results from running the algorithm on the data post induction.Top: Activation profile plot of iModulon 34.Bottom: Genes weights above threshold for IModulon 34.

Fig
Fig S53| Overall identification of iModulons based on the heavy metal with highest aggregate coefficient in the activation window.

Fig
Fig S54| Classification of each iModulon described in Figure 2 and 3 according to the shape of the activity coefficients within the induction window of heavy metal enriched.