DZAP1 L-curve analysis KH

L-curve analysis

Github issue

Github issue 172

Purpose

We are attempting to determine which alpha value we should use when running the model.

Parameters

For all input sheets the optimization parameters were set:

• alpha : 1.00E-02
• kk_max : 1
• MaxIter : 1.00E+08
• TolFun : 1.00E-10
• MaxFunEval : 1.00E+08
• TolX : 1.00E-10
• production_function : Sigmoid
• L_curve : 1.00E+00
• estimate_params : 1.00E+00
• make_graphs : 0.00E+00
• fix_P : 0
• fix_b : 0

Running the L-curve

• with_deletions_ONLY_DNA_binding_dZAP1_15_Genes_2_1_2016
  • Computer: Run on Seaver 120-12
  • Code Version: GRNmap beta 6:32 pm 2/1/2016
  • Has 27 edges
  • Run started: 6:31 pm 2/1/2016
  • The program finished running at 8:37 am 2/2/2016
  • There were 16 input xlsx, output xlsx, and output.mat files created
  • The input sheets (and thus the output sheets) were saved in a folder called “KristenHorstmann” in the thaw space. A file called KristenHorstmannwith_deletions_ONLY_DNA_binding_dZAP1_15_Genes_2_1_2016LCurveLCurveData was output and saved in the thaw space (not in the “KristenHorstmann” folder
  • There is only one optimization_diagnostic.jpg, meaning that likely it is overwriting after every run.

• with_deletions_ONLY_DNA_binding_dZAP1_15_Genes_2_1_2016
  • Computer: Run on Seaver 120-13
  • Code Version: GRNmap beta 6:38 pm 2/1/2016
  • Run started: 6:40 pm 2/1/2016
  • The program computer 6 alphas and crashed on the 7th alpha.
  • The error was:

Attempted to access strain_x1(2,:); index out of bounds because numel(strain_x1)=1. Error in lse (line 104) x1 = strain_x1(i,:); Error in GRNLCurve (line 66) GRNstruct = lse(GRNstruct); Error in GRNmodel (line 32) GRNstruct = GRNLCurve(GRNstruct);

• • Accidentally was the same sheet run on computer 12, ran new one as stated below:

• with_deletions_ONLY_DNA_34_Genes_2_1
  • Computer: Run on SEA 120-13
  • Code Version: GRNmap beta 6:38 pm 2/1/2016
  • Run started: 5:42 pm 2/2/2016
  • By the wa7, has 98 edges, likely why taking so long.
  • The program was still running at 4:13 PM 2/3/16 and was on the 5th alpha, which started at 2:58 AM.
  • The program was still running at 6:26 PM 2/4/16 and was on the 7th alpha, which started at 7:05 AM.
  • Program was still running at 2:40 PM 2/5/16 and was on the 8th alpha.

• ONLY_DNA_binding_dZAP1_28_genes_2_1_16
  • Computer: Run on Seaver 120-14
  • Code Version: GRNmap beta 6:35 pm 2/1/2016
  • Has 43 edges
  • Run started: 7:38 pm 2/1/2016
  • The program was still running at 8:56 am 2/2/2016. At this time it was on the 8th alpha.
  • Program was sitll running at 4:15 PM 2/3/16. At this time it was on the 12th alpha, which started at 12:38 AM.
  • The program was still running at 6:27 PM 2/4/16 and was on the 14th alpha, which started at 9:59 AM.
  • Program was still running at 2:43 PM on 2/5 and was on the 16th alpha.

• ONLY_DNA_binding_dZAP1_14_genes_2_1
  • Has 15 edges
  • Computer: Run on Seaver 120-15
  • Code Version: GRNmap beta 6:34 pm 2/1/2016
  • Run started: 7:45 pm 2/1/2016
  • The program finished running at 3:11 am 2/2/2016
  • There were 16 input xlsx, output xlsx, and output.mat files created
  • The input sheets (and thus the output sheets) were saved in a folder called “KHorstmann” in the thaw space. A file called KHorstmannONLY_DNA_binding_dZAP1_14_genes_2_1LCurveLCurveData was output and saved in the thaw space (not in the “KHorstmann” folder.
  • There is only one optimization_diagnostic.jpg, meaning that likely it is overwriting after every run.

Output of L-Curve Analysis

• Excel sheet of the optimization diagnostics combined to create L-curve plots: Media:KH L curves small networks.xlsx
• Powerpoint of the L-curves with alpha labeling: Media:ZAP1 L curve analysis graphs KH 02-03-2016.pptx
• The l-curves are slightly off for mine vs Tessa's. Neither line is exceptionally smooth; DNA_ONLY_binding_dZAP1_14_genes has a small spike in the middle of the line, while with_deletions_DNA_ONLY_binding_dZAP1_15_genes has a large dip towards the end of the alpha count, which skews the rest of the y-axis
• Checked that data was being copied from correct sheets and that all lines were copied and pasted to the correct alpha.
• Ran l-curve of 10-allpha outputs of 33 gene network: Media:ZAP1 Lcurve analysis 33 genes 98 edges.pdf
  • Last four alphas were all stacked on top of each other for some reason did not vary muxh
• Attempted to fix L-curve graphs of 33 genes and 25 genes, but data still very "stacked" towards the end, making the expression data unreadable.
  • Media:ZAP1_l-curve_33-genes_fullalphas.pdf
  • Media:ZAP1_l-curve_25-genes_fixed_2-19.pdf