Recent changes to advanced_tutorial

advanced_tutorial modified by Wei Li

Wei Li — Wed, 11 Nov 2020 17:14:22 -0000

--- v27
+++ v28
@@ -258,7 +258,7 @@
 Instead using the naive approach in option 1, users may want to get the genes that have differences in later time points (weeks 3,4) vs. early time points (weeks 1,2), as well as genes that show differences in drug treatment vs. DMSO controls. The design matrix can be set up as follows.

-sample|baseline|early|latevsearly|E2
+sample|baseline|common|latevsearly|E2
 -----|-|-|-|-
 day0|1|0|0|0
 dmso-1wk|1|1|0|0
@@ -274,7 +274,7 @@

 beta score|meaning
 --------|--------
-early| The corresponding gene, if this score is nonzero, has phenotype in early time points (weeks 1-2), **DMSO** samples. Negative score means this gene is essential.
+common| The corresponding gene, if this score is nonzero, has common signals across all the samples. In particular, this column should capture signals in early time points (weeks 1-2), **DMSO** samples, as "common" factor is the only non-zero factor for these samples. Negative score means this gene is essential.
 latevsearly|The corresponding gene,  if this score is nonzero, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples. Negative score means this gene becomes more essential in late time points.
 E2|The corresponding gene, if this score is nonzero, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 18:56:51 -0000

--- v26
+++ v27
@@ -212,7 +212,7 @@
 You can type "mageck" to test if it works.

-## Tutorial 4: make full use of MAGeCK MLE for more complicated experimental design (e.g., time series) ##
+## Tutorial 4: make full use of MAGeCK MLE for more complicated experimental design (e.g., paired samples, time series) ##

 Besides simple design matrix (i.e., one "beta score" for one type of sample), users can design more complicated design matrix for their needs including taking time series CRISPR screens into consideration.

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 18:56:14 -0000

--- v25
+++ v26
@@ -303,13 +303,14 @@
 time| The linear factor for the corresponding genes in time series, **assuming the log fold change is linear**. if the abs(time  beta score) is high, this gene has consistent changes in **DMSO** samples. Negative score means this gene is essential.
 E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.

-We tested this model in our T47D CRISPR screening dataset (E2 vs DMSO, weeks 1-4). The results of two example genes are shown below.
+We tested this model in our CRISPR screening dataset (E2 vs DMSO, weeks 1-4) on an ER+ breast cancer cell line (T47D). The results of two example genes are shown below.

 Gene|time_beta|E2_beta
 ------|-----|-----
 CSK|0.463|-0.58
 FOXA1|0.027|-0.465

+These results demonstrated CSK is a tumor suppressor gene shown in DMSO samples (score=0.463) but the function is somehow reverted in E2 (E2 score=-0.58), consistent with our [previous findings](https://www.pnas.org/content/115/31/7869.short). In contrast, FOXA1, a known oncogene in Estrogen and Estrogen Receptor signaling, has little effect on DMSO (score=0.027) but is depleted in E2 treated cells (score=-0.465).


 ### Considering paired sample information in MLE ###

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 18:51:56 -0000

--- v24
+++ v25
@@ -303,6 +303,15 @@
 time| The linear factor for the corresponding genes in time series, **assuming the log fold change is linear**. if the abs(time  beta score) is high, this gene has consistent changes in **DMSO** samples. Negative score means this gene is essential.
 E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.

+We tested this model in our T47D CRISPR screening dataset (E2 vs DMSO, weeks 1-4). The results of two example genes are shown below.
+
+Gene|time_beta|E2_beta
+------|-----|-----
+CSK|0.463|-0.58
+FOXA1|0.027|-0.465
+
+
+
 ### Considering paired sample information in MLE ###

 Note that in MAGeCK-RRA, users can add a *--paired*  option to take paired sample information into consideration. In the MLE module, users can easily set up design matrix to consider paired sample. For example, users want to directly compare treatments vs controls, and the experiment is performed on two paired samples, where (treatment_rep1, control_rep1) and (treatment_rep2, control_rep2) are performed independently.

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 18:44:09 -0000

--- v23
+++ v24
@@ -300,10 +300,33 @@

 beta score|meaning
 --------|--------
-time| The linear factor for the corresponding genes in time series, **assuming the change is linear**. if the abs(time  beta score) is high, this gene has consistent changes in **DMSO** samples. Negative score means this gene is essential.
+time| The linear factor for the corresponding genes in time series, **assuming the log fold change is linear**. if the abs(time  beta score) is high, this gene has consistent changes in **DMSO** samples. Negative score means this gene is essential.
 E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.

-
+### Considering paired sample information in MLE ###
+
+Note that in MAGeCK-RRA, users can add a *--paired*  option to take paired sample information into consideration. In the MLE module, users can easily set up design matrix to consider paired sample. For example, users want to directly compare treatments vs controls, and the experiment is performed on two paired samples, where (treatment_rep1, control_rep1) and (treatment_rep2, control_rep2) are performed independently.
+
+In MAGeCK-RRA, the corresponding command line would be
+
+    :::bash
+    mageck test -t treatment_rep1,treatment_rep2 -c control_rep1,control_rep2 --paired
+    
+In MAGeCK-VISPR, the corresponding design matrix would be 
+
+sample|baseline|t_vs_c|rep2_vs_rep1
+-----|-|-|-
+control_rep1|1|0|0
+control_rep2|1|0|1
+treatment_rep1|1|1|0
+treatment_rep2|1|1|1
+
+The meanings of the scores are as follows.
+
+beta score|meaning
+--------|--------
+t_vs_c| The beta score differences between treatment vs control.
+rep2_vs_rep1|This beta score models the differences between rep 1 and rep 2.


 ## Tutorial 5: Include the sgRNA efficiency into mle calculation ##

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 18:28:46 -0000

--- v22
+++ v23
@@ -215,6 +215,8 @@
 ## Tutorial 4: make full use of MAGeCK MLE for more complicated experimental design (e.g., time series) ##

 Besides simple design matrix (i.e., one "beta score" for one type of sample), users can design more complicated design matrix for their needs including taking time series CRISPR screens into consideration. 
+
+### A time-series drug treatment example ###

 One particular example is given below. Genome-wide CRISPR screens are performed on a 4-week time series on cells treated with one drug and control (DMSO). In the end, users get the following samples:

@@ -233,7 +235,7 @@

 Users can have multiple ways to set up the design matrix.

-### Option 1: a simple design matrix ###
+#### Option 1: a simple design matrix ####

 Users can design one beta score (i.e., one column in the design matrix) for each sample type (1-4 weeks * drug type):

@@ -251,7 +253,7 @@

 In the end, MAGeCK-VISPR will output eight beta scores (and their associated p values), each beta score corresponds to one of the eight samples (dmso-1wk,e2-1wk,dmso-2wk,e2-2wk,dmso-3wk,e2-3wk,dmso-4wk,e2-4wk). Further analysis can be performed (e.g., clustering) based on these beta scores

-### Option 2: consider early vs late time points, and drug treatment differences ###
+#### Option 2: consider early vs late time points, and drug treatment differences ####

 Instead using the naive approach in option 1, users may want to get the genes that have differences in later time points (weeks 3,4) vs. early time points (weeks 1,2), as well as genes that show differences in drug treatment vs. DMSO controls. The design matrix can be set up as follows.

@@ -272,11 +274,11 @@

 beta score|meaning
 --------|--------
-early| The corresponding genes, if the abs(early  beta score) is high, have phenotype in early time points (weeks 1-2), **DMSO** samples. Negative score means this gene is essential.
-latevsearly|The corresponding genes,  if the abs(latevsearly  beta score) is high, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples. Negative score means this gene becomes more essential in late time points.
-E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.
-
-### Option 3: linear time points and drug treatments ###
+early| The corresponding gene, if this score is nonzero, has phenotype in early time points (weeks 1-2), **DMSO** samples. Negative score means this gene is essential.
+latevsearly|The corresponding gene,  if this score is nonzero, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples. Negative score means this gene becomes more essential in late time points.
+E2|The corresponding gene, if this score is nonzero, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.
+
+#### Option 3: linear time points and drug treatments ####

 Assuming genes may behave linearly in time points, users may design a matrix like the following.

@@ -292,7 +294,7 @@
 dmso-4wk|1|**4**|0
 e2-4wk|1|**4**|1

-Note that we can have other positive numbers than "1" in the design matrix. Relatively, if a log fold change of a gene in week 1 is x, the expected log fold change of that gene in week 2 (and 3,4) would be 2x (3x, 4x), etc. Note that this simplified scenario may not hold true in many cases.  
+Note that we can have other positive numbers than "1" in the design matrix. Relatively, if a log fold change of a gene in week 1 is x, the expected log fold change of that gene in week 2 (and 3,4) would be 2x (3x, 4x), etc.  This assumes a linear reduction (or enrichment) of the cell population on a log scale; for example, a cell with certain gene knockout reduces 50% of its population every week. Therefore the log2 fold change for weeks 1-4 would be -1, -2, -3, -4.  *Note that this simplified linear scenario may not hold true in many cases. * 

 In the end, MAGeCK-VISPR will output two beta scores (and their associated p values). The meanings are described below.

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 16:54:50 -0000

--- v21
+++ v22
@@ -251,7 +251,7 @@

 In the end, MAGeCK-VISPR will output eight beta scores (and their associated p values), each beta score corresponds to one of the eight samples (dmso-1wk,e2-1wk,dmso-2wk,e2-2wk,dmso-3wk,e2-3wk,dmso-4wk,e2-4wk). Further analysis can be performed (e.g., clustering) based on these beta scores

-### Option 2: consider early or late time points of drug treatment differences ###
+### Option 2: consider early vs late time points, and drug treatment differences ###

 Instead using the naive approach in option 1, users may want to get the genes that have differences in later time points (weeks 3,4) vs. early time points (weeks 1,2), as well as genes that show differences in drug treatment vs. DMSO controls. The design matrix can be set up as follows.

@@ -272,9 +272,35 @@

 beta score|meaning
 --------|--------
-early| The corresponding genes, if the abs(early  beta score) is high, have phenotype in early time points (weeks 1-2), **DMSO** samples.
-latevsearly|The corresponding genes,  if the abs(latevsearly  beta score) is high, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples
-E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples.
+early| The corresponding genes, if the abs(early  beta score) is high, have phenotype in early time points (weeks 1-2), **DMSO** samples. Negative score means this gene is essential.
+latevsearly|The corresponding genes,  if the abs(latevsearly  beta score) is high, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples. Negative score means this gene becomes more essential in late time points.
+E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.
+
+### Option 3: linear time points and drug treatments ###
+
+Assuming genes may behave linearly in time points, users may design a matrix like the following.
+
+sample|baseline|time|E2
+-----|-|-|-
+day0|1|0|0
+dmso-1wk|1|1|0
+e2-1wk|1|1|1
+dmso-2wk|1|**2**|0
+e2-2wk|1|**2**|1
+dmso-3wk|1|**3**|0
+e2-3wk|1|**3**|1
+dmso-4wk|1|**4**|0
+e2-4wk|1|**4**|1
+
+Note that we can have other positive numbers than "1" in the design matrix. Relatively, if a log fold change of a gene in week 1 is x, the expected log fold change of that gene in week 2 (and 3,4) would be 2x (3x, 4x), etc. Note that this simplified scenario may not hold true in many cases.  
+
+In the end, MAGeCK-VISPR will output two beta scores (and their associated p values). The meanings are described below.
+
+beta score|meaning
+--------|--------
+time| The linear factor for the corresponding genes in time series, **assuming the change is linear**. if the abs(time  beta score) is high, this gene has consistent changes in **DMSO** samples. Negative score means this gene is essential.
+E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples. Negative score means this gene is more essential within E2 drug treatment.
+

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 16:44:49 -0000

--- v20
+++ v21
@@ -233,7 +233,7 @@

 Users can have multiple ways to set up the design matrix.

-### Option 1: a simple design matrix ##
+### Option 1: a simple design matrix ###

 Users can design one beta score (i.e., one column in the design matrix) for each sample type (1-4 weeks * drug type):

@@ -251,7 +251,7 @@

 In the end, MAGeCK-VISPR will output eight beta scores (and their associated p values), each beta score corresponds to one of the eight samples (dmso-1wk,e2-1wk,dmso-2wk,e2-2wk,dmso-3wk,e2-3wk,dmso-4wk,e2-4wk). Further analysis can be performed (e.g., clustering) based on these beta scores

-## Option 2: consider early or late time points of drug treatment differences ##
+### Option 2: consider early or late time points of drug treatment differences ###

 Instead using the naive approach in option 1, users may want to get the genes that have differences in later time points (weeks 3,4) vs. early time points (weeks 1,2), as well as genes that show differences in drug treatment vs. DMSO controls. The design matrix can be set up as follows.

@@ -268,6 +268,14 @@
 dmso-4wk|1|1|1|0
 e2-4wk|1|1|1|1

+In the end, MAGeCK-VISPR will output three beta scores (and their associated p values). The meanings are described below.
+
+beta score|meaning
+--------|--------
+early| The corresponding genes, if the abs(early  beta score) is high, have phenotype in early time points (weeks 1-2), **DMSO** samples.
+latevsearly|The corresponding genes,  if the abs(latevsearly  beta score) is high, shows different behavior in late time points (weeks 3-4) vs. early time points in **DMSO** samples
+E2|The corresponding genes, if the abs(E2 beta score) is high, shows difference in **drug treated(E2)** samples vs. DMSO samples.
+


 ## Tutorial 5: Include the sgRNA efficiency into mle calculation ##

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 16:39:54 -0000

--- v19
+++ v20
@@ -244,7 +244,31 @@
 e2-1wk|1|0|1|0|0|0|0|0|0
 dmso-2wk|1|0|0|1|0|0|0|0|0
 e2-2wk|1|0|0|0|1|0|0|0|0
-        
+dmso-3wk|1|0|0|0|0|1|0|0|0
+e2-3wk|1|0|0|0|0|0|1|0|0
+dmso-4wk|1|0|0|0|0|0|0|1|0
+e2-4wk|1|0|0|0|0|0|0|0|1
+
+In the end, MAGeCK-VISPR will output eight beta scores (and their associated p values), each beta score corresponds to one of the eight samples (dmso-1wk,e2-1wk,dmso-2wk,e2-2wk,dmso-3wk,e2-3wk,dmso-4wk,e2-4wk). Further analysis can be performed (e.g., clustering) based on these beta scores
+
+## Option 2: consider early or late time points of drug treatment differences ##
+
+Instead using the naive approach in option 1, users may want to get the genes that have differences in later time points (weeks 3,4) vs. early time points (weeks 1,2), as well as genes that show differences in drug treatment vs. DMSO controls. The design matrix can be set up as follows.
+
+
+sample|baseline|early|latevsearly|E2
+-----|-|-|-|-
+day0|1|0|0|0
+dmso-1wk|1|1|0|0
+e2-1wk|1|1|0|1
+dmso-2wk|1|1|0|0
+e2-2wk|1|1|0|1
+dmso-3wk|1|1|1|0
+e2-3wk|1|1|1|1
+dmso-4wk|1|1|1|0
+e2-4wk|1|1|1|1
+
+

 ## Tutorial 5: Include the sgRNA efficiency into mle calculation ##

advanced_tutorial modified by Wei Li

Wei Li — Fri, 09 Oct 2020 16:30:46 -0000

--- v18
+++ v19
@@ -231,7 +231,20 @@
 e2-4wk|4 weeks|E2

-
+Users can have multiple ways to set up the design matrix.
+
+### Option 1: a simple design matrix ##
+
+Users can design one beta score (i.e., one column in the design matrix) for each sample type (1-4 weeks * drug type):
+
+sample|baseline|dmso-1wk|e2-1wk|dmso-2wk|e2-2wk|dmso-3wk|e2-3wk|dmso-4wk|e2-4wk
+-----|-|-|-|-|-|-|-|-|-
+day0|1|0|0|0|0|0|0|0|0
+dmso-1wk|1|1|0|0|0|0|0|0|0
+e2-1wk|1|0|1|0|0|0|0|0|0
+dmso-2wk|1|0|0|1|0|0|0|0|0
+e2-2wk|1|0|0|0|1|0|0|0|0
+        

 ## Tutorial 5: Include the sgRNA efficiency into mle calculation ##