CSL-Tox

This tutorial contains the necessary scripts and data to reproduce the work explained in the paper “CSL-Tox: An open-source analytical framework for the comparison of short-term and long-term toxicity end points and exploring the opportunities for decreasing in-vivo studies conducted for drug development programs.”

Naga D and Dimitrakopoulou S et al.

An overview of the steps performed and implemented in the CSL-Tox workflow.

1- Data refinement

• Uploading necessary libraries and scripts

library(tidyverse)
library(purrr)
#importing scripts containing functions for reading the data and transform the dataset
source("ReadingData.R")
source("ChangeDataset.R")
options(knitr.kable.NA = '')

• Assigning the file names. The main and pathology datasets are provided in the Data folder: Tables S3 and S4 respectively.

MainData_file="Data/S3.txt" 
PathologyData_file="Data/S4.txt"

• Reading the data

Data=ReadingData(MainData_file,PathologyData_file)

• Refining the dataset:

1. Grouping all species into rodents & non rodents
2. Apply controled terminology
3. Group findings into the six high level defined categories described in the paper :

High level categories
body_weight
body_weight_gain
git_clinical_signs
other_clinical_signs
macroscopic
pathology
organ_weight
cardiovascular_effects

Data=ChangeData(Data)

2- Data exploration

• Viewing the data

knitr::kable(head(Data$MainData))

study_id	identifier	therapeutic_area	type	route_of_administration	species	duration	dose	adj_dose	pre_treatment	dose_interval	dose_lethality	noael	noael_sex	adverse_findings	body_weight	body_weight_gain	death_diagnosis	neurological_clinical_signs	git_clinical_signs	other_clinical_signs	lab_measurements	macroscopic	pathology	organ_weight	cardiovascular_effects	group
Study 1	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	non-rodent	39	0	0.00		bi weekly		no		no	no	no	no	no	no	no	no	no	no	no	no	long
Study 1	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	non-rodent	39	5	0.36		bi weekly		no		no	no	no	no	no	no	no	no	no	no	no	no	long
Study 1	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	non-rodent	39	25	1.79		bi weekly		no		no	no	no	no	no	no	no	no	no	no	no	no	long
Study 1	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	non-rodent	39	100	7.14		bi weekly		yes	both	no	no	yes	no	no	no	no	no	no	no	no	no	long
Study 2	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	rodent	26	0	0.00		bi weekly		no		no	no	no	no	no	no	no	no	no	no	no	no	long
Study 2	Compound 1	Cardiovascular and Metabolism	sm	subcutaneous	rodent	26	2.5	0.18		bi weekly		no		no	yes	yes	no	yes	yes	yes	yes-r	yes	yes	no	no	long

knitr::kable(summary(Data))

	Length	Class	Mode
MainData	27	data.frame	list
body_weight_Dataset	14	data.frame	list
neurological_clinical_signs_Dataset	14	data.frame	list
git_clinical_signs_Dataset	14	data.frame	list
other_clinical_signs_Dataset	14	data.frame	list
macroscopic_Dataset	15	data.frame	list
organ_weight_Dataset	15	data.frame	list
cardiovascular_effects_Dataset	14	data.frame	list
pathology_Dataset	15	data.frame	list

• Plotting an overview of the data:

#importing the script containing the functions
source("DataExploratoryPlots.R")

Therapeutic_Areas.fn(Data$MainData)

Different_Species.fn(Data$MainData)

Dist.St.fn(Data$MainData)

StudiesPerFindingCat(Data$MainData)

 Dist.St.fn(Data$MainData)

3- Data analysis

This section reproduces the main results included in the paper: Overall adversity of molecules, NOAEL changes and likelihood ratios.

• Calculation of the “Appearance” dataframe : Table containing the drug,modality,species, findings and whether those findings were observed “1” or not observed “0” in short, middle and long studies

#importing the script containing the functions
source("Appearance_Functions.R")
Appearance=map_dfr(.x =unique(Data$MainData$identifier),.f = Appear_Drug, Data)

#Viewing the head of the appearance table 
knitr::kable(head(Appearance))

drug	modality	species	finding	short	middle	long
Compound 1	sm	rodent	Weight changes	1	1	1
Compound 1	sm	non-rodent	Weight changes	1	0	1
Compound 1	sm	rodent	Neurological clinical signs	1	0	1
Compound 1	sm	non-rodent	Neurological clinical signs	0	0	0
Compound 1	sm	rodent	Gastrointestinal clinical signs	0	0	1
Compound 1	sm	non-rodent	Gastrointestinal clinical signs	0	0	0

• Refining the appearance dataframe: Grouping short and middle study together

Summarised.Appearance = Cond.Appearance.fn(Appearance)
#head of summarized appearance table
knitr::kable(head(Summarised.Appearance))

drug	modality	species	finding	short.cond	middle.cond	long.cond	short_middle.cond
Compound 1	sm	non-rodent	Cardiovascular System	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Cutaneous	TRUE	FALSE	TRUE	TRUE
Compound 1	sm	non-rodent	Endocrine System	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Exocrine System	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Eye/conjuctiva	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Gastrointestinal clinical signs	FALSE	FALSE	FALSE	FALSE

Note: The finding coloumn in the summarized appearance table contains the high level terms, for more detailed findings the following function can be used:

Summarised.Appearance.detailed = Cond.Appearance.ext.fn(Appearance)
knitr::kable(head(Summarised.Appearance.detailed))

drug	modality	species	finding	short.cond	middle.cond	long.cond	short_middle.cond
Compound 1	sm	non-rodent	Cardiovascular System/ microscopic pathology	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Cardiovascular System/macroscopic pathology	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Cardiovascular System/organ weight change	FALSE	FALSE	FALSE	FALSE
Compound 1	sm	non-rodent	Cutaneous/ microscopic pathology	TRUE	FALSE	TRUE	TRUE
Compound 1	sm	non-rodent	Cutaneous/macroscopic pathology	TRUE	FALSE	TRUE	TRUE
Compound 1	sm	non-rodent	Endocrine System/ microscopic pathology	FALSE	FALSE	FALSE	FALSE

• Repeating previous analysis but for adverse events only

Adverse.Data=AdverseData.fn(Data)
Adverse.Appearance=map_dfr(.x =unique(Adverse.Data$MainData$identifier),.f = Appear_Drug, Adverse.Data)
Summarised.Adverse.Appearance=Cond.Appearance.fn(Adverse.Appearance)

(i) Calculation of the overall adversity per modality

• For SM

adversity_sm = Adversity.Summary.fn(Summarised.Adverse.Appearance, type="sm")
knitr::kable(adversity_sm, caption = "Adversity summary for small molecules")

drug	short_middle	long
Compound 1	FALSE	FALSE
Compound 2	FALSE	FALSE
Compound 3	FALSE	FALSE
Compound 4	FALSE	FALSE
Compound 5	FALSE	FALSE
Compound 6	FALSE	TRUE
Compound 7	FALSE	TRUE
Compound 10	TRUE	FALSE
Compound 11	TRUE	FALSE
Compound 8	TRUE	FALSE
Compound 9	TRUE	FALSE
Compound 12	TRUE	TRUE
Compound 13	TRUE	TRUE
Compound 14	TRUE	TRUE
Compound 15	TRUE	TRUE
Compound 16	TRUE	TRUE
Compound 17	TRUE	TRUE
Compound 18	TRUE	TRUE
Compound 19	TRUE	TRUE
Compound 20	TRUE	TRUE
Compound 21	TRUE	TRUE
Compound 22	TRUE	TRUE
Compound 23	TRUE	TRUE
Compound 24	TRUE	TRUE
Compound 25	TRUE	TRUE

Adversity summary for small molecules

• For LM

adversity_lm = Adversity.Summary.fn(Summarised.Adverse.Appearance, type="lm")
knitr::kable(adversity_lm, caption = "Adversity summary for large molecules")

drug	short_middle	long
Compound a	FALSE	FALSE
Compound b	FALSE	FALSE
Compound c	FALSE	FALSE
Compound d	FALSE	FALSE
Compound e	FALSE	FALSE
Compound f	FALSE	FALSE
Compound g	FALSE	FALSE
Compound h	FALSE	FALSE
Compound i	FALSE	FALSE
Compound j	FALSE	FALSE
Compound k	FALSE	FALSE
Compound l	FALSE	FALSE
Compound m	FALSE	FALSE
Compound n	FALSE	FALSE
Compound o	FALSE	TRUE
Compound p	TRUE	TRUE
Compound q	TRUE	TRUE
Compound r	TRUE	TRUE

Adversity summary for large molecules

(ii) Calculation of the NOAEL changes from short to long-term studies

#importing the script containing the functions
source("NOAEL_Change.R")
#table with results for sm
knitr::kable(result_sm, caption = "Noael changes for small molecules")

identifier	noael	species
Compound 1	Increase	non-rodent
Compound 2	Same	non-rodent
Compound 3	Increase	non-rodent
Compound 4	Same	non-rodent
Compound 5	Decrease	non-rodent
Compound 6	Decrease	non-rodent
Compound 8	Increase	non-rodent
Compound 9	Same	non-rodent
Compound 10	Same	non-rodent
Compound 11	Increase	non-rodent
Compound 12	Decrease	non-rodent
Compound 13	Decrease	non-rodent
Compound 14	Decrease	non-rodent
Compound 15	Increase	non-rodent
Compound 17	Same	non-rodent
Compound 18	Decrease	non-rodent
Compound 19	Same	non-rodent
Compound 20	Decrease	non-rodent
Compound 21	Increase	non-rodent
Compound 22	Increase	non-rodent
Compound 24	Decrease	non-rodent
Compound 25	Decrease	non-rodent
Compound 1	Same	rodent
Compound 2	Same	rodent
Compound 3	Same	rodent
Compound 4	Decrease	rodent
Compound 5	Decrease	rodent
Compound 6	Same	rodent
Compound 9	Same	rodent
Compound 10	Decrease	rodent
Compound 11	Increase	rodent
Compound 12	Same	rodent
Compound 13	Same	rodent
Compound 14	Decrease	rodent
Compound 15	Increase	rodent
Compound 16	Same	rodent
Compound 17	Same	rodent
Compound 18	Decrease	rodent
Compound 19	Increase	rodent
Compound 20	Decrease	rodent
Compound 21	Same	rodent
Compound 22	Decrease	rodent
Compound 23	Decrease	rodent
Compound 24	Increase	rodent
Compound 25	Decrease	rodent

Noael changes for small molecules

#table with results for lm
knitr::kable(result_lm,caption = "Noael changes for large molecules")

identifier	noael	species
Compound a	Same	non-rodent
Compound b	Same	non-rodent
Compound c	Increase	non-rodent
Compound d	Decrease	non-rodent
Compound e	Decrease	non-rodent
Compound f	Increase	non-rodent
Compound g	Increase	non-rodent
Compound h	Increase	non-rodent
Compound i	Increase	non-rodent
Compound j	Same	non-rodent
Compound k	Increase	non-rodent
Compound l	Increase	non-rodent
Compound m	Same	non-rodent
Compound n	Same	non-rodent
Compound o	Decrease	non-rodent
Compound q	Same	non-rodent
Compound r	Decrease	non-rodent
Compound d	Same	rodent
Compound k	Increase	rodent
Compound m	Same	rodent
Compound n	Same	rodent
Compound p	Decrease	rodent
Compound q	Decrease	rodent

Noael changes for large molecules

(iii) Calculation of the Likelihood ratios

#importing the script containing the functions
source("LikelihoodRatio.R")

• Calculate contingency tables and likelihood ratios for all findings

Specify rodent or non-rodent:

animal = "rodent"
Likelihood.Ratio=Likelihood.Ratio.fn(Summarised.Appearance %>% filter(species== animal))

• Round values and select only significant likelihood ratios with p-values < 0.05

Likelihood.Ratio[,-1]=round(Likelihood.Ratio[,-1],2)

#Likelihood ratios with p.value<0.05
Likelihood.Ratio.imp=Likelihood.Ratio %>% filter(p_value<=0.05) %>% 
  arrange(desc(LR_pos), desc(iLR_neg)) %>% select(-c(Sensitivity,Specificity))

knitr::kable(Likelihood.Ratio.imp,caption = "Significant Likelihood ratios for rodents for all findings")

finding	TP	FP	FN	TN	LR_pos	iLR_neg	p_value
Reproductive System	4	1	2	23	16.00	2.87	0.00
Gastrointestinal clinical signs	7	2	2	19	8.17	4.07	0.00
GI tract	5	2	4	19	5.83	2.04	0.01
Cutaneous	5	3	3	19	4.58	2.30	0.02
Endocrine System	13	4	1	12	3.71	10.50	0.00
liver	14	4	2	10	3.06	5.71	0.00
Lymphoid Tissues	9	5	4	12	2.35	2.29	0.04
Weight changes	14	4	4	8	2.33	3.00	0.02

Significant Likelihood ratios for rodents for all findings

• Plot frequency of fp and fn across the findings in rodent and non-rodent

Appear_plot(dataf = Likelihood.Ratio.imp ,species = "rodent", legend_pos=c(0.90,0.90))

• Calculate contingency tables and likelihood ratios for adverse findings only (using the adverse appearance table this time)

Adverse.Likelihood.Ratio=Likelihood.Ratio.fn(Summarised.Adverse.Appearance %>% filter(species== animal))

• Round values and select only significant likelihood ratios with p-values < 0.05

Adverse.Likelihood.Ratio[,-1]=round(Adverse.Likelihood.Ratio[,-1],2)

#Likelihood ratios with p.value<0.05
Adverse.Likelihood.Ratio.imp=Adverse.Likelihood.Ratio %>% filter(p_value<=0.05) %>% 
arrange(desc(LR_pos), desc(iLR_neg)) %>%  select(-c(Sensitivity,Specificity))

knitr::kable(Adverse.Likelihood.Ratio,caption = "Significant Likelihood ratios for rodents for adverse findings in rodents")

finding	TP	FP	FN	TN	Sensitivity	Specificity	LR_pos	iLR_neg	p_value
Cardiovascular System	0	1	3	26	0.00	0.96	0.00	0.96	1.00
Cutaneous	0	4	2	24	0.00	0.86	0.00	0.86	1.00
Endocrine System	2	4	1	23	0.67	0.85	4.50	2.56	0.09
Exocrine System	1	1	4	24	0.20	0.96	5.00	1.20	0.31
Eye/conjuctiva	0	0	1	29	0.00	1.00		1.00	1.00
Gastrointestinal clinical signs	0	1	0	29		0.97			1.00
GI tract	2	3	1	24	0.67	0.89	6.00	2.67	0.06
In life cardiovascular effects	0	0	0	30		1.00			1.00
liver	3	0	1	26	0.75	1.00	Inf	4.00	0.00
Lymphoid Tissues	2	2	1	25	0.67	0.93	9.00	2.78	0.04
MuscularSkeletal System	0	3	1	26	0.00	0.90	0.00	0.90	1.00
Nervous System	1	0	3	26	0.25	1.00	Inf	1.33	0.13
Neurological clinical signs	2	2	2	24	0.50	0.92	6.50	1.85	0.07
Other clinical signs	1	4	2	23	0.33	0.85	2.25	1.28	0.43
Reproductive System	0	2	2	26	0.00	0.93	0.00	0.93	1.00
Respiratory System	0	1	1	28	0.00	0.97	0.00	0.97	1.00
Urinary System	0	1	5	24	0.00	0.96	0.00	0.96	1.00
Weight changes	0	6	3	21	0.00	0.78	0.00	0.78	1.00

Significant Likelihood ratios for rodents for adverse findings in rodents

• Plot frequency of fp and fn across the adverse findings

Appear_plot(dataf = Adverse.Likelihood.Ratio.imp ,species = "rodent", legend_pos=c(0.90,0.90))

The previous steps are repeated to generate the results for the non-rodent species.