added docker and code for PDX data build

Author: sgosline

build/docker/Dockerfile.mpnstPDX

@@ -0,0 +1,26 @@
+FROM r-base:4.3.2
+ENV DEBIAN_FRONTEND=noninteractive
+RUN apt-get update --allow-insecure-repositories
+#RUN apt-get install -y --allow-unauthenticated build-essential --fix-missing libpq-dev python3-pip python3-setuptools python3-dev python3-venv libcurl4-openssl-dev libxml2-dev
+RUN apt-get install -y --allow-unauthenticated build-essential --fix-missing python3-pip python3-setuptools python3-dev python3-venv libcurl4-openssl-dev libxml2-dev 
+
+RUN python3 -m venv /opt/venv
+RUN /opt/venv/bin/pip3 install --upgrade pip
+
+
+ENV PYTHONPATH "${PYTHONPATH}:/app"
+WORKDIR /app
+
+ADD build/mpnstPDX/requirements.txt .
+ADD build/mpnstPDX/requirements.r .
+ADD build/mpnstPDX/* ./
+ADD build/utils/* ./
+
+# installing python libraries
+RUN /opt/venv/bin/pip3 install -r requirements.txt
+
+# installing r libraries
+RUN Rscript requirements.r
+
+
+VOLUME ["/tmp"]

build/mpnstPDX/02_get_drug_data.R

@@ -101,7 +101,7 @@ ignore_file_path <- '/tmp/mpnstpdx_ignore_chems.txt'
 use_python("/opt/venv/bin/python3", required = TRUE)
 source_python("pubchem_retrieval.py")
 
-update_dataframe_and_write_tsv(unique_names=alldrugs,output_filename=output_file_path,ignore_chems=ignore_file_path)
+update_dataframe_and_write_tsv(unique_names=drugs,output_filename=output_file_path,ignore_chems=ignore_file_path)
 
 
 tab<-read.table(newdrugfile,sep='\t',header=T,quote="",comment.char="")

build/mpnstPDX/03_get_drug_response_data.R

@@ -97,6 +97,7 @@ pdx_meta$parentId=unlist(lapply(pdx_meta$id,function(x) synGet(x)$parentId))
 oldfolders=c('syn22018363','syn22024460','syn22024428','syn22024429','syn22024437','syn22024438')
 
 old_meta<-subset(pdx_meta,parentId%in%oldfolders)
+
 old_data<-do.call(rbind,lapply(unique(old_meta$parentId),function(x){
     ids<-subset(old_meta,parentId==x)|>
         subset(!is.na(id))
@@ -121,7 +122,7 @@ old_data<-do.call(rbind,lapply(unique(old_meta$parentId),function(x){
        }))
 }))|>
   left_join(unique(select(old_meta,id=parentId,improve_sample_id)))|>
-  dplyr::select(experiment=id,model_id=improve_sample_id,specimen_id,treatment=chem_name,time=experimental_time_point,volume=assay_value)|>distinct()
+  dplyr::select(experiment=id,model_id=improve_sample_id,specimen_id,treatment=chem_name,time=experimental_time_point,time_unit=experimental_time_point_unit,volume=assay_value)|>distinct()
 
 
 
@@ -150,7 +151,7 @@ new_data<-do.call(rbind,lapply(unique(new_meta$id), function(x){
     #print(head(tab))
     return(tab)}))|>
     left_join(pdx_meta)|>
-    dplyr::select(experiment=id,model_id=improve_sample_id,specimen_id,treatment=chem_name,time=experimental_time_point,volume=assay_value)|>distinct()
+    dplyr::select(experiment=id,model_id=improve_sample_id,specimen_id,treatment=chem_name,time=experimental_time_point,time_unit=experimental_time_point_unit,volume=assay_value)|>distinct()
 
 ##maybe tweak the data frame a bit depending on curve fitting script
 pdx_data<-rbind(old_data,new_data)
@@ -161,7 +162,7 @@ pdx_data<-rbind(old_data,new_data)
 fwrite(pdx_data,'/tmp/curve_data.tsv',sep='\t')
 
 ##TODO: create new curve fitting script in python
-pycmd = '/opt/venv/bin/python fit_pdx_curve.py --input /tmp/curve_data.tsv --output /tmp/mpnstPDX_experiments.tsv'
+pycmd = '/opt/venv/bin/python calc_pdx_metrics.py --input /tmp/curve_data.tsv --outprefix /tmp/mpnstPDX'
 print('running curve fitting')
 #system(pycmd)
 

build/mpnstPDX/build_exp.sh

@@ -1,2 +1,2 @@
 Rscript 03_get_drug_response_data.R $SYNAPSE_AUTH_TOKEN $1 $2
-/opt/venv/bin/python3 compute_metrics.py /tmp/file.tsv --drugfile=/tmp/mpnstpdx_drugs.tsv
+/opt/venv/bin/python3 compute_metrics.py /tmp/curve_data.tsv --drugfile=/tmp/mpnstpdx_drugs.tsv --outprefix=/tmp/mpnstpdx

build/utils/calc_pdx_metrics.py

@@ -187,9 +187,9 @@ def AUC(time, volume, time_normalize=True):
     #print(time)
     if time_normalize:
         auc = auc/np.max(time)
-    return {"metric": "auc", "value": auc}
+    return {"metric": "auc", "value": auc, 'time':np.max(time)}
 
-def TGI(contr_volume, treat_volume):
+def TGI(contr_volume, treat_volume,time):
     """
     Computes the tumor growth inhibition (TGI) between two time-volume curves.
 
@@ -207,7 +207,8 @@ def TGI(contr_volume, treat_volume):
     # Simulated batch response class object
     rtx = {
         "metric": "TGI",
-        "value": tgi
+        "value": tgi,
+        'time': np.max(time)
     }
     return rtx
 
@@ -240,7 +241,7 @@ def ABC(contr_time=None, contr_volume=None, treat_time=None, treat_volume=None):
         tre = AUC(treat_time, treat_volume)
 
     abc = con['value'] - tre['value']
-    return {"metric": "abc", "value": abc}#, "control": con, "treatment": tre}
+    return {"metric": "abc", "value": abc,'time':np.max(treat_time)}#, "control": con, "treatment": tre}
 
 
 ###LMM CODE
@@ -292,13 +293,14 @@ def lmm(time, volume, treatment, drug_name):
         'interaction_coef_value': i_coef_value
     }
 
-    return {'metric': 'lmm','value': i_coef_value}
+    return {'metric': 'lmm','value': i_coef_value,'time':np.max(time)}
 
 def main():
     parser=ArgumentParser()
     ###read in file with model id, volume, time, condition
     parser.add_argument('curvefile')
     parser.add_argument('--drugfile')
+    parser.add_argument('--outprefix',default='/tmp/')
 
     args = parser.parse_args()
 
@@ -308,8 +310,24 @@ def main():
     singles, combos = get_drug_stats(tab)
 
     ##join with drug ids
+    expsing = singles.rename({'drug':'chem_name','metric':'drug_combination_metric','value':'drug_combination_value','sample':'improve_sample_id'},axis=1).merge(drugs,on='chem_name',how='left')[['improve_drug_id','improve_sample_id','drug_combination_metric','drug_combination_value']]
+    expsing = expsing.dropna()
+
+    combos[['drug1','drug2']]=combos.drug.str.split('+',expand=True)
+    combos = combos.rename({'metric':'drug_combination_metric','value':'drug_combination_value','sample':'improve_sample_id'},axis=1).dropna()
+
+    expcomb = combos.rename({'drug1':'chem_name'},axis=1).merge(drugs,on='chem_name',how='left').rename({'improve_drug_id':'improve_drug_1'},axis=1)[['improve_drug_1','drug2','improve_sample_id','drug_combination_metric','drug_combination_value']]
+    expcomb = expcomb.rename({'drug2':'chem_name'},axis=1).merge(drugs,on='chem_name',how='left').rename({'improve_drug_id':'improve_drug_2'},axis=1)[['improve_drug_1','improve_drug_2','improve_sample_id','drug_combination_metric','drug_combination_value']]
+
+    expcomb[['source']]='Synapse'
+    expcomb[['study']]='MPNST PDX in vivo'
+
+    expsing[['source']]='Synapse'
+    expsing[['study']]='MPNST PDX in vivo'
+    expsing.to_csv(args.outprefix+'_experiments.csv',index=False)
+    expcomb.to_csv(args.outprefix+'_combinations.csv',index=False)
 
-    ##stats.to_csv('pdx_exp.tsv',sep='\t',index=False)
+
 
 def get_drug_stats(df,control='control'):
     ##for each experiment, call group
@@ -341,24 +359,24 @@ def get_drug_stats(df,control='control'):
             treat_abc = ABC(ctl_time,ctl_volume,treat_time,treat_volume)
             #print(f"AUC: {treat_auc}")
             #print(f"ABC: {treat_abc}")
-            treat_abc.update({'sample':mod,'drug':d})
+            treat_abc.update({'sample':mod,'drug':d,'time_unit':'days'})
             if '+' in d:
                 combores.append(treat_abc)
             else:
                 singleres.append(treat_abc)
             #lmm
             comb = pd.concat([ctl_data,d_data])
             lmm_res = lmm(comb.time, comb.volume, comb.treatment,d)
-            lmm_res.update({'sample':mod,'drug':d})
+            lmm_res.update({'sample':mod,'drug':d,'time_unit':'days'})
             #print(f"LMM: {lmm_res}")
             if '+' in d:
-                combores.append(lm_res)
+                combores.append(lmm_res)
             else:
-                singleres.append(lm_res)
+                singleres.append(lmm_res)
 
             #get tgi for group
-            tg = TGI(ctl_volume,treat_volume)
-            tg.update({'sample':mod,'drug':d})
+            tg = TGI(ctl_volume,treat_volume,treat_time)
+            tg.update({'sample':mod,'drug':d,'time_unit':'days'})
             #print(tg)
             if '+' in d:
                 combores.append(tg)
@@ -368,13 +386,13 @@ def get_drug_stats(df,control='control'):
 
             #get mRECIST for group
             mr = mrecist(treat_time,treat_volume)
-            mr.update({'sample':mod,'drug':d})
+            mr.update({'sample':mod,'drug':d,'time_unit':'days'})
             if '+' in d:
                 combores.append(mr)
             else:
                 singleres.append(mr)
 
-    sing = pd.DataFrame.from_records(singlres)
+    sing = pd.DataFrame.from_records(singleres)
     comb = pd.DataFrame.from_records(combores)
     return sing,comb
 

schema/coderdata.yaml

@@ -187,13 +187,13 @@ classes:
       - source
       - study
     attributes:
-      drug_one_id:
+      drug_drug_2:
         description: improve_drug_id of first drug
-      drug_two_id:
+      improve_drug_2:
         description: imrrove_drug_id of second drug
-      combination_metric:
+      drug_combination_metric:
         description: metric calculated for synergy, or other metric of two drugs
-      combination_value:
+      drug_combination_value:
         description: value of metric for synergy or combination
 enums:
   ResponseMetric: