Pipeline-description.md

@@ -64,7 +64,29 @@ The last _ab initio_ gene prediction is done on the PN40024 assemblies with [Gen
 
 Combining all the predictions generated with the different tools at the preceding step.
 
-### Filtering
+### Filtering (AGAT)
+
+The gene models from EvidenceModeler are filtered with the following rules : 
+
+Ab initio supported gene models were kept if :
+  * there are predicted by at least 2 ab initio predictors
+  * if the start and stop condon was present
+  * if gene length >= 300 bp
+Ab initio supported gene models not matching these constraints were kept if :
+  * they had a database hit with the Uniprot/SwissProt or NCBI NR database (blastp hits with en evalue <1e-6)
+Gene models only supported by evidence data or by lifted annotation were kept if :
+  * if the start and stop condon was present
+  * if gene length >= 300 bp
+
+At then end, we have the final gene models (final gff3).
+
+### PASA
+
+Used to add UTR regions in the final GFF3.
+
+### tRNAscan-SE
+
+Used to annotation tRNA
 
 ## Data