11# Depreciated rules that may still be useful for some projects
2+ def get_normal_sorted_bam (wildcards ):
3+ """
4+ Returns a tumor samples paired normal
5+ See config['pairs'] for tumor, normal pairs.
6+ """
7+ normal = tumor2normal [wildcards .name ]
8+ if normal :
9+ # Runs in a tumor, normal mode
10+ return join (workpath , "BAM" , "{0}.sorted.bam" .format (normal ))
11+ else :
12+ # Runs in tumor-only mode
13+ return []
14+
215
316# Depreciated germline variant calling rule(s)
417rule deepvariant :
@@ -57,4 +70,83 @@ rule deepvariant:
5770 --output_vcf={output.vcf} \\
5871 --num_shards={threads} \\
5972 --intermediate_results_dir=${{tmp}}
60- """
73+ """
74+
75+ # Depreciated somatic variant calling rule(s)
76+ rule deepsomatic :
77+ """
78+ Data processing step to call somatic variants using deep neural
79+ network in tumor-normal pairs. DeepSomatic is an extension of the
80+ deep learning-based variant caller DeepVariant that takes aligned
81+ reads (in BAM or CRAM format) from tumor and normal data, produces
82+ pileup image tensors from them, classifies each tensor using a CNN,
83+ and finally reports somatic variants in a standard VCF or gVCF file.
84+ This rule runs all three steps in the deepsomatic pipeline as a one
85+ step: i.e. make_examples, call_variants, and postprocess_variants.
86+ This is not optimal for large-scale projects as it will consume a lot
87+ of resources inefficently (only the 2nd step in the dv pipeline can
88+ make use of GPU-computing). As so, it is better to run the 1st/3rd
89+ step on a normal compute node and run the 2nd step on a GPU node.
90+ @Input:
91+ Duplicate marked, sorted Tumor-Normal BAM file (scatter)
92+ @Output:
93+ Single-sample VCF file with called somatic variants
94+ """
95+ input :
96+ tumor = join (workpath , "BAM" , "{name}.sorted.bam" ),
97+ normal = get_normal_sorted_bam
98+ output :
99+ vcf = join (workpath , "deepsomatic" , "somatic" , "{name}.deepsomatic.vcf" ),
100+ params :
101+ rname = "deepsom" ,
102+ genome = config ['references' ]['GENOME' ],
103+ tmpdir = tmpdir ,
104+ # Building option for deepsomatic config, where:
105+ # @WGS = --model_type=WGS
106+ # @WES = --model_type=WES (may be added in future)
107+ dv_model_type = "WGS" ,
108+ # Get tumor and normal sample names
109+ tumor = '{name}' ,
110+ # Building option for the paired normal sorted bam
111+ normal_bam_option = lambda w : "--reads_normal={0}.sorted.bam" .format (
112+ join (workpath , "BAM" , tumor2normal [w .name ])
113+ ) if tumor2normal [w .name ] else "" ,
114+ # Building option for the normal sample name
115+ normal_name_option = lambda w : "--sample_name_normal={0}" .format (
116+ tumor2normal [w .name ]
117+ ) if tumor2normal [w .name ] else "" ,
118+ threads : int (allocated ("threads" , "deepsomatic" , cluster ))
119+ container : config ['images' ]['deepsomatic' ]
120+ envmodules : config ['tools' ]['deepsomatic' ]
121+ shell : """
122+ # Setups temporary directory for
123+ # intermediate files with built-in
124+ # mechanism for deletion on exit
125+ if [ ! -d "{params.tmpdir}" ]; then mkdir -p "{params.tmpdir}"; fi
126+ tmp=$(mktemp -d -p "{params.tmpdir}")
127+ trap 'du -sh "${{tmp}}"; rm -rf "${{tmp}}"' EXIT
128+
129+ # Export OpenBLAS variable to
130+ # control the number of threads
131+ # in a thread pool. By setting
132+ # this variable to 1, work is
133+ # done in the thread that ran
134+ # the operation, rather than
135+ # disbatching the work to a
136+ # thread pool. If this option
137+ # is not provided, it can lead
138+ # to nested parallelism.
139+ # See this issue for more info:
140+ # https://github.com/google/deepsomatic/issues/28
141+ export OPENBLAS_NUM_THREADS=1
142+
143+ # Run deepsomatic
144+ run_deepsomatic \\
145+ --model_type={params.dv_model_type} \\
146+ --ref={params.genome} \\
147+ --reads_tumor={input.tumor} {params.normal_bam_option} \\
148+ --sample_name_tumor={params.tumor} {params.normal_name_option} \\
149+ --output_vcf={output.vcf} \\
150+ --num_shards={threads} \\
151+ --intermediate_results_dir=${{tmp}}
152+ """
0 commit comments