From 5af15a1c1d4061003d39b7475573ccce3ac6a0e5 Mon Sep 17 00:00:00 2001
From: Jennifer Chang <jennifer.chang.bioinform@gmail.com>
Date: Fri, 24 May 2024 10:40:46 -0700
Subject: [PATCH] Add a configurable includes.txt

Adds a configurable includes_{serotype}.txt to force-include key strains
(e.g. vaccine-related, lineage-defining) for each serotype tree.
---
 phylogenetic/config/config_dengue.yaml   | 1 +
 phylogenetic/config/include_all.txt      | 1 +
 phylogenetic/config/include_denv1.txt    | 1 +
 phylogenetic/config/include_denv2.txt    | 1 +
 phylogenetic/config/include_denv3.txt    | 1 +
 phylogenetic/config/include_denv4.txt    | 1 +
 phylogenetic/rules/prepare_sequences.smk | 2 ++
 7 files changed, 8 insertions(+)
 create mode 100644 phylogenetic/config/include_all.txt
 create mode 100644 phylogenetic/config/include_denv1.txt
 create mode 100644 phylogenetic/config/include_denv2.txt
 create mode 100644 phylogenetic/config/include_denv3.txt
 create mode 100644 phylogenetic/config/include_denv4.txt

diff --git a/phylogenetic/config/config_dengue.yaml b/phylogenetic/config/config_dengue.yaml
index a110c969..ebd9a318 100644
--- a/phylogenetic/config/config_dengue.yaml
+++ b/phylogenetic/config/config_dengue.yaml
@@ -9,6 +9,7 @@ display_strain_field: "strain"
 
 filter:
   exclude: "config/exclude.txt"
+  include: "config/include_{serotype}.txt"
   group_by: "year region"
   min_length:
     genome: 5000
diff --git a/phylogenetic/config/include_all.txt b/phylogenetic/config/include_all.txt
new file mode 100644
index 00000000..b98b661e
--- /dev/null
+++ b/phylogenetic/config/include_all.txt
@@ -0,0 +1 @@
+# Format: <GenBank accession> [# <documentation on reason for inclusion>]
\ No newline at end of file
diff --git a/phylogenetic/config/include_denv1.txt b/phylogenetic/config/include_denv1.txt
new file mode 100644
index 00000000..b98b661e
--- /dev/null
+++ b/phylogenetic/config/include_denv1.txt
@@ -0,0 +1 @@
+# Format: <GenBank accession> [# <documentation on reason for inclusion>]
\ No newline at end of file
diff --git a/phylogenetic/config/include_denv2.txt b/phylogenetic/config/include_denv2.txt
new file mode 100644
index 00000000..b98b661e
--- /dev/null
+++ b/phylogenetic/config/include_denv2.txt
@@ -0,0 +1 @@
+# Format: <GenBank accession> [# <documentation on reason for inclusion>]
\ No newline at end of file
diff --git a/phylogenetic/config/include_denv3.txt b/phylogenetic/config/include_denv3.txt
new file mode 100644
index 00000000..b98b661e
--- /dev/null
+++ b/phylogenetic/config/include_denv3.txt
@@ -0,0 +1 @@
+# Format: <GenBank accession> [# <documentation on reason for inclusion>]
\ No newline at end of file
diff --git a/phylogenetic/config/include_denv4.txt b/phylogenetic/config/include_denv4.txt
new file mode 100644
index 00000000..b98b661e
--- /dev/null
+++ b/phylogenetic/config/include_denv4.txt
@@ -0,0 +1 @@
+# Format: <GenBank accession> [# <documentation on reason for inclusion>]
\ No newline at end of file
diff --git a/phylogenetic/rules/prepare_sequences.smk b/phylogenetic/rules/prepare_sequences.smk
index 7f651b68..60144f85 100644
--- a/phylogenetic/rules/prepare_sequences.smk
+++ b/phylogenetic/rules/prepare_sequences.smk
@@ -55,6 +55,7 @@ rule filter:
         sequences = lambda wildcard: "data/sequences_{serotype}.fasta" if wildcard.gene in ['genome'] else "results/{gene}/sequences_{serotype}.fasta",
         metadata = "data/metadata_{serotype}.tsv",
         exclude = config["filter"]["exclude"],
+        include = config["filter"]["include"],
     output:
         sequences = "results/{gene}/filtered_{serotype}.fasta"
     params:
@@ -69,6 +70,7 @@ rule filter:
             --metadata {input.metadata} \
             --metadata-id-columns {params.strain_id} \
             --exclude {input.exclude} \
+            --include {input.include} \
             --output {output.sequences} \
             --group-by {params.group_by} \
             --sequences-per-group {params.sequences_per_group} \