diff --git a/config/cluster/slurm.json b/config/cluster/slurm.json
index 0627f77..ead9ed5 100644
--- a/config/cluster/slurm.json
+++ b/config/cluster/slurm.json
@@ -290,6 +290,11 @@
         "time" : "1-00:00:00",
         "threads": "24"
     },
+    "clairs_tumor_only": {
+        "mem": "64G",
+        "time" : "1-00:00:00",
+        "threads": "24"
+    },
     "somatic_purple_maf": {
         "threads": "4",
         "mem": "32G",
diff --git a/config/cluster/uge.json b/config/cluster/uge.json
index e8e4a6e..6d0937e 100644
--- a/config/cluster/uge.json
+++ b/config/cluster/uge.json
@@ -281,6 +281,11 @@
     "partition": "",
     "threads": "8"
   },
+  "clairs_tumor_only": {
+    "mem": "4G",
+    "partition": "",
+    "threads": "8"
+  },
   "inbreeding": {
     "mem": "4G",
     "partition": "",
diff --git a/config/containers.json b/config/containers.json
index e57119f..19c7fae 100644
--- a/config/containers.json
+++ b/config/containers.json
@@ -10,6 +10,7 @@
         "base":                 "docker://skchronicles/ccbr_wes_base:v0.1.0",
         "deepvariant_gpu":      "docker://google/deepvariant:1.5.0-gpu",
         "deepvariant":          "docker://google/deepvariant:1.5.0",
+        "clairs-to":            "docker://hkubal/clairs-to:v0.2.0",
         "glnexus":              "docker://ghcr.io/dnanexus-rnd/glnexus:v1.4.1",
         "open_cravat":          "docker://skchronicles/ncbr_opencravat:v0.1.0",
         "octopus":              "docker://skchronicles/ncbr_octopus:v0.2.0",
@@ -17,4 +18,4 @@
         "sequenza":             "docker://sequenza/sequenza:3.0.0",
         "vcf2maf":              "docker://skchronicles/ncbr_vcf2maf:v0.1.0"
     }
-}
+}
\ No newline at end of file
diff --git a/workflow/rules/somatic.smk b/workflow/rules/somatic.smk
index a18c153..1436334 100644
--- a/workflow/rules/somatic.smk
+++ b/workflow/rules/somatic.smk
@@ -690,6 +690,62 @@ rule hmftools_sage:
     """
 
 
+rule clairs_tumor_only:
+    """Data-processing step to call somatic variants in tumor-only samples using 
+    ClairS. ClairS is a deep-learning based variant caller that uses an ensembl
+    of two neural networks to call somatic variants. ClairS-TO is unique in that
+    it can call somatic variants without a matched normal. More information about 
+    ClairS-TO can be found here: https://github.com/HKU-BAL/ClairS-TO
+    @Input:
+        Realigned, recalibrated BAM file (scatter-per-tumor-sample)
+    @Output:
+        Per sample somatic variants in VCF format
+    """
+    input:
+        tumor = join(workpath, "BAM", "{name}.recal.bam"),
+    output:
+        snps   = join(workpath, "clairs", "somatic", "{name}", "snv.vcf.gz"),
+        indels = join(workpath, "clairs", "somatic", "{name}", "indel.vcf.gz"),
+        tmp    = join(workpath, "clairs", "somatic", "{name}", "clairs_snps_indels.vcf"),
+        vcf    = join(workpath, "clairs", "somatic", "{name}.clairs.vcf"),
+    params:
+        rname   = 'clairs_to',
+        tumor   = '{name}',
+        genome  = config['references']['GENOME'],
+        outdir = join(workpath, "clairs", "somatic", "{name}"),
+    threads: 
+        int(allocated("threads", "clairs_tumor_only", cluster)),
+    container: config['images']['clairs-to']
+    envmodules: config['tools']['rlang']
+    shell: """
+    # Call somatic variants with ClairS-TO,
+    # run in isolated sample directory to
+    # collisions in file names
+    /opt/bin/run_clairs_to \\
+        --tumor_bam_fn {input.tumor}  \\
+        --ref_fn {params.genome} \\
+        --threads {threads} \\
+        --platform  ilmn \\
+        --output_dir {params.outdir} \\
+        --conda_prefix /opt/micromamba/envs/clairs-to
+    
+    # Concatenate SNPs and Indels
+    bcftools concat \\
+        -a \\
+        -O v \\
+        -o {output.tmp} \\
+        {output.snps} \\
+        {output.indels} 
+    
+    # Filter for PASS variants
+    bcftools view \\
+        -f 'PASS' \\
+        -O v \\
+        -o {output.vcf} \\
+        {output.tmp}
+    """
+
+
 rule muse:
     """Data-processing step to call somatic mutations with MuSE. This tool is 
     unique in accounting for tumor heterogeneity using a sample-specific error