ONT_pilot_w_GDB_samples.md

@@ -652,6 +652,49 @@ cd /scratch/users/sbusi/ONT/cedric_ont_basecalling/2019_GDB/bins
 dRep compare drep_output -g ./*.fa
 ```
 
+##### Assessing MAGs #####
+- Although expected, as stated somewhere above, "we are scientists and so, we need evidence!"
+- I decided to confirm that the bins recovered via each of the assembly methods, and additionally the mapping methods would be different
+- To get there, first we needed to link the CheckM and GTDBtk outputs to the binned MAGs
+```
+# Working directory
+cd /scratch/users/sbusi/ONT/cedric_ont_basecalling/2019_GDB/results/Binning/checkm_output
+
+# making a list of all the samples
+ls -1 *.txt | sed 's/_output.txt//g' > sample_list
+
+# collating the checkm_output to a readable format
+for file in `cat sample_list`
+do
+  sed '1,/Bin Id/d;/INFO/,$d' "$file"_output.txt | \
+  sed '/^-----/d' | \
+    awk '{print $1"\t"$13"\t"$14"\t"$15}' | \
+      sed $'1 i\\\nbin\tcompletion\tcontamination\tstrain_heterogeneity' > "$file"_checkm.txt 
+done
+
+# collecting the gtdbtk information
+for file in `cat sample_list`
+do
+  awk '{print $1"\t"$2}' ../gtdbtk_output/"$file"/classify/gtdbtk.bac120.summary.tsv | sed 's/user_genome/bin/g' > "$file"_gtdbtk.txt
+done
+
+# merging files based on a common first column
+for file in `cat sample_list`
+do
+  awk -F"\t" 'BEGIN{OFS="\t"} {if (NR==FNR) {a[$1]=$2"\t"$3"\t"$4; next} if ($1 in a) {print $1, $2, a[$1]}}' \
+    "$file"_checkm.txt "$file"_gtdbtk.txt > "$file"_taxonomy_contamination.txt 
+done
+
+# keeping only those bins with completion over 60% and contamination below 10% (this number can be adjusted as needed)
+for file in `cat sample_list`
+do
+  awk '$3>=60 && $4 <10 {print}' "$file"_taxonomy_contamination.txt > "$file"_HQ_bins.txt
+done
+
+# The "$file"_HQ_bins.txt files were subsequently analysed on the Desktop in R using the "MAGS_2019_GDB_analyses.R" script
+```
+
+
 ##### VizBin #####
 - The binners did great for the sr- and hybrid assemblies, however, fared poorly with the lr- (flye) assembly
 - A sign of a good researcher is that they never trust their own data, and validate it twelve-ways-to-Sunday