Re add nextclade assignment rules

ingest/Snakefile

@@ -57,6 +57,7 @@ rule all:
 
 include: "workflow/snakemake_rules/fetch_sequences.smk"
 include: "workflow/snakemake_rules/transform.smk"
+include: "workflow/snakemake_rules/nextclade.smk"
 
 
 if config.get("upload", False):

ingest/workflow/snakemake_rules/nextclade.smk

@@ -0,0 +1,178 @@
+rule nextclade_all:
+    input:
+        sequences="results/sequences.fasta",
+        dataset="../nextclade_data/all",
+    output:
+        "results/nextclade_results/nextclade_all.tsv",
+    threads: 4
+    shell:
+        """
+        nextclade run \
+          --input-dataset {input.dataset} \
+          -j {threads} \
+          --output-tsv {output} \
+          --min-match-rate 0.01 \
+          --silent \
+          {input.sequences}
+        """
+
+# Might be able to parallelize this
+rule split_dengue_sequences:
+    input:
+        sequences="results/sequences.fasta",
+        metadata="results/metadata.tsv",
+        nextclade_all_results="results/nextclade_results/nextclade_all.tsv",
+    output:
+        sequences_denv1="results/sequences_denv1.fasta",
+        sequences_denv2="results/sequences_denv2.fasta",
+        sequences_denv3="results/sequences_denv3.fasta",
+        sequences_denv4="results/sequences_denv4.fasta",
+    shell:
+        """
+        augur filter \
+          --sequences results/sequences.fasta \
+          --metadata {input.nextclade_all_results} \
+          --metadata-id-columns seqName \
+          --query "clade=='DENV1'" \
+          --output-sequences {output.sequences_denv1}
+        
+        augur filter \
+          --sequences results/sequences.fasta \
+          --metadata {input.nextclade_all_results} \
+          --metadata-id-columns seqName \
+          --query "clade=='DENV2'" \
+          --output-sequences {output.sequences_denv2}
+
+        augur filter \
+          --sequences results/sequences.fasta \
+          --metadata {input.nextclade_all_results} \
+          --metadata-id-columns seqName \
+          --query "clade=='DENV3'" \
+          --output-sequences {output.sequences_denv3}
+
+        augur filter \
+          --sequences results/sequences.fasta \
+          --metadata {input.nextclade_all_results} \
+          --metadata-id-columns seqName \
+          --query "clade=='DENV4'" \
+          --output-sequences {output.sequences_denv4}
+        """
+
+# The following rules should use wildcards
+rule nextclade_denv1:
+    input:
+        sequences="results/sequences_denv1.fasta",
+        dataset="../nextclade_data/denv1",
+    output:
+        "results/nextclade_results/nextclade_denv1.tsv",
+    threads: 4
+    shell:
+        """
+        nextclade run \
+          --input-dataset {input.dataset} \
+          -j {threads} \
+          --output-tsv {output} \
+          --min-match-rate 0.01 \
+          --silent \
+          {input.sequences}
+        """
+
+rule nextclade_denv2:
+    input:
+        sequences="results/sequences_denv2.fasta",
+        dataset="../nextclade_data/denv2",
+    output:
+        "results/nextclade_results/nextclade_denv2.tsv",
+    threads: 4
+    shell:
+        """
+        nextclade run \
+          --input-dataset {input.dataset} \
+          -j {threads} \
+          --output-tsv {output} \
+          --min-match-rate 0.01 \
+          --silent \
+          {input.sequences}
+        """
+
+rule nextclade_denv3:
+    input:
+        sequences="results/sequences_denv3.fasta",
+        dataset="../nextclade_data/denv3",
+    output:
+        "results/nextclade_results/nextclade_denv3.tsv",
+    threads: 4
+    shell:
+        """
+        nextclade run \
+          --input-dataset {input.dataset} \
+          -j {threads} \
+          --output-tsv {output} \
+          --min-match-rate 0.01 \
+          --silent \
+          {input.sequences}
+        """
+
+rule nextclade_denv4:
+    input:
+        sequences="results/sequences_denv4.fasta",
+        dataset="../nextclade_data/denv4",
+    output:
+        "results/nextclade_results/nextclade_denv4.tsv",
+    threads: 4
+    shell:
+        """
+        nextclade run \
+          --input-dataset {input.dataset} \
+          -j {threads} \
+          --output-tsv {output} \
+          --min-match-rate 0.01 \
+          --silent \
+          {input.sequences}
+        """
+
+rule join_nextclade_clades:
+    input:
+        metadata="results/metadata.tsv",
+        nextclade_denv1="results/nextclade_results/nextclade_denv1.tsv",
+        nextclade_denv2="results/nextclade_results/nextclade_denv2.tsv",
+        nextclade_denv3="results/nextclade_results/nextclade_denv3.tsv",
+        nextclade_denv4="results/nextclade_results/nextclade_denv4.tsv",
+    output:
+        metadata_all="results/metadata_all.tsv",
+        metadata_denv1="results/metadata_denv1.tsv",
+        metadata_denv2="results/metadata_denv2.tsv",
+        metadata_denv3="results/metadata_denv3.tsv",
+        metadata_denv4="results/metadata_denv4.tsv",
+    shell:
+        """
+        echo "genbank_accession,nextclade_subtype,nextclade_type" \
+        | tr ',' '\t' \
+        > results/nextclade_subtype.tsv
+
+        tsv-select -H -f "seqName,clade" {input.nextclade_denv1} \
+        | awk 'NR>1 {{print $0"\tDENV1"}}' \
+        >> results/nextclade_subtype.tsv
+        tsv-select -H -f "seqName,clade" {input.nextclade_denv2} \
+        | awk 'NR>1 {{print $0"\tDENV2"}}' \
+        >> results/nextclade_subtype.tsv
+        tsv-select -H -f "seqName,clade" {input.nextclade_denv3} \
+        | awk 'NR>1 {{print $0"\tDENV3"}}' \
+        >> results/nextclade_subtype.tsv
+        tsv-select -H -f "seqName,clade" {input.nextclade_denv4} \
+        | awk 'NR>1 {{print $0"\tDENV4"}}' \
+        >> results/nextclade_subtype.tsv
+
+        tsv-join -H \
+            --filter-file results/nextclade_subtype.tsv \
+            --key-fields genbank_accession \
+            --append-fields 'nextclade_subtype,nextclade_type' \
+            --write-all ? \
+            {input.metadata} \
+        > {output.metadata_all}
+
+        tsv-filter -H --str-eq nextclade_type:DENV1 {output.metadata_all} > {output.metadata_denv1}
+        tsv-filter -H --str-eq nextclade_type:DENV2 {output.metadata_all} > {output.metadata_denv2}
+        tsv-filter -H --str-eq nextclade_type:DENV3 {output.metadata_all} > {output.metadata_denv3}
+        tsv-filter -H --str-eq nextclade_type:DENV4 {output.metadata_all} > {output.metadata_denv4}
+        """