diff --git a/.dockerignore b/.dockerignore
new file mode 100644
index 0000000..ef985a5
--- /dev/null
+++ b/.dockerignore
@@ -0,0 +1,5 @@
+# Ignore everything
+*
+
+# Allow requirements file
+!/requirements.txt
diff --git a/Dockerfile b/Dockerfile
index 6d7428f..91418d8 100644
--- a/Dockerfile
+++ b/Dockerfile
@@ -1,14 +1,13 @@
-FROM python:3.11-slim-bullseye
+FROM python:3.11.4-slim-bookworm
 
 # Install prerequisites
 RUN apt-get update && \
 apt-get upgrade --assume-yes && \
 apt-get install --assume-yes --no-install-recommends wget
 
-# ADD bin/install_pplacer.sh /tmp/install_pplacer.sh
-# RUN /tmp/install_pplacer.sh
-
-RUN pip install pandas csvkit fastalite
+WORKDIR /usr/local/src/yapp/
+ADD requirements.txt .
+RUN pip install --requirement requirements.txt
 
 # create some mount points
 RUN mkdir -p /app /fh /mnt /run/shm
diff --git a/SConstruct b/SConstruct
index 04cd8c9..6e1bafd 100644
--- a/SConstruct
+++ b/SConstruct
@@ -104,7 +104,7 @@ to_remove = input.get('to_remove')
 singularity = conf['singularity'].get('singularity', 'singularity')
 deenurp_img = conf['singularity']['deenurp']
 dada2_img = conf['singularity']['dada2']
-csvkit_img = conf['singularity']['csvkit']
+epa_img = conf['singularity']['epa']
 yapp_img = conf['singularity']['yapp']
 
 binds = [os.path.abspath(pth)
@@ -130,17 +130,10 @@ vars.Add('venv', None, venv)
 # paths here to avoid accidental introduction of external
 # dependencies.
 
-# find the execution path for singularity; assumes 'ml Singularity' has been run
-try:
-    singularity_bin = [pth for pth in os.environ['PATH'].split(':')
-                       if 'singularity' in pth][0]
-except IndexError:
-    sys.exit('PATH for Singularity not found: try running\nml Singularity')
-
 env = SlurmEnvironment(
     ENV=dict(
         os.environ,
-        PATH=':'.join(['bin', path.join(venv, 'bin'), singularity_bin,
+        PATH=':'.join(['bin', path.join(venv, 'bin'),
                        '/usr/local/bin', '/usr/bin', '/bin']),
         SLURM_ACCOUNT='fredricks_d',
         OMP_NUM_THREADS=args.nproc),
@@ -153,7 +146,7 @@ env = SlurmEnvironment(
     binds=' '.join('-B {}'.format(pth) for pth in ['$cwd'] + binds),
     deenurp_img=('$singularity exec $binds --pwd $cwd {}'.format(deenurp_img)),
     dada2_img=('$singularity exec $binds --pwd $cwd {}'.format(dada2_img)),
-    csvkit_img=('$singularity exec $binds --pwd $cwd {}'.format(csvkit_img)),
+    epa_img=('$singularity exec $binds --pwd $cwd {}'.format(epa_img)),
     yapp_img=('$singularity exec $binds --pwd $cwd {}'.format(yapp_img)),
     min_reads=min_reads,
 )
@@ -173,6 +166,7 @@ for_transfer = [settings]
 
 profile = common.taxit_rp(refpkg, 'profile', img=deenurp_img, singularity=singularity)
 ref_sto = common.taxit_rp(refpkg, 'aln_sto', img=deenurp_img, singularity=singularity)
+ref_tree = common.taxit_rp(refpkg, 'tree', img=deenurp_img, singularity=singularity)
 
 # filter non-16s reads with cmsearch
 seqs_16s, seqs_not16s, cmsearch_scores = env.Command(
@@ -221,11 +215,28 @@ merged, = env.Command(
             'rm ${TARGET}.temp')
 )
 
+reference = env.Command(
+    target='$out/refs_merged.afa',
+    source=[merged, ref_sto],
+    action='unmerge.py --stockholm ${SOURCES[1]} --out $TARGET ${SOURCES[0]}')
+
+query = env.Command(
+    target='$out/query_merged.afa',
+    source=[merged, query_sto],
+    action='unmerge.py --stockholm ${SOURCES[1]} --out $TARGET ${SOURCES[0]}')
+
 dedup_jplace, = env.Command(
-    target='$out/dedup.jplace',
-    source=[refpkg, merged],
-    action=('$deenurp_img pplacer -p --inform-prior --prior-lower 0.01 --map-identity '
-            '-c $SOURCES -o $TARGET -j $nproc'),
+    target='$out/epa_result.jplace',
+    source=[query, reference, ref_tree],
+    action='$epa_img epa-ng '
+            # '--filter-max 999999999 '  # all placements
+            '--model GTR+G  '
+            '--out-dir $out '
+            '--query ${SOURCES[0]} '
+            '--redo '
+            '--ref-msa ${SOURCES[1]} '
+            '--threads $nproc '
+            '--tree ${SOURCES[2]}'
     # ncores=nproc,
     # slurm_queue=large_queue
 )
@@ -239,10 +250,11 @@ classify_db, = env.Command(
     action=('rm -f $TARGET && '
             '$deenurp_img rppr prep_db -c ${SOURCES[1]} --sqlite $TARGET && '
             '$deenurp_img guppy classify '
-            '--pp --classifier hybrid2 '  # TODO: specify pplacer settings in config
+            '--classifier hybrid2 '  # TODO: specify pplacer settings in config
             '-j $nproc '
             '${SOURCES[0]} '  # placefile
             '-c ${SOURCES[1]} '
+            '--nbc-rank species '
             '--nbc-sequences ${SOURCES[2]} '
             '--sqlite $TARGET ')
 )
diff --git a/bin/get_classifications.py b/bin/get_classifications.py
index 5e34083..5c5c7c1 100755
--- a/bin/get_classifications.py
+++ b/bin/get_classifications.py
@@ -213,7 +213,7 @@ def main(arguments):
         try:
             new_tax_id, __, __ = tax.primary_from_name(tax_name)
         except ValueError:
-            log.error(f'could not find {tax_name} in taxonomy')
+            log.error('could not find {} in taxonomy'.format(tax_name))
             missing_tax_names.append(tax_name)
         else:
             new_tax_ids.append(new_tax_id)
diff --git a/bin/jplace.py b/bin/jplace.py
new file mode 100644
index 0000000..1bbd773
--- /dev/null
+++ b/bin/jplace.py
@@ -0,0 +1,27 @@
+from Bio import Phylo
+
+
+class JParser(Phylo.NewickIO.Parser):
+    def _parse_tree(self, text):
+        tree = super()._parse_tree(text)
+        return Phylo.PhyloXML.Phylogeny(
+            root=tree.root,
+            rooted=tree.rooted)
+
+    def new_clade(sef, parent=None):
+        clade = JClade()
+        if parent:
+            clade.parent = parent
+        return clade
+
+
+class JClade(Phylo.PhyloXML.Clade):
+    def __init__(self, edge=None, **kwds):
+        Phylo.PhyloXML.Clade.__init__(self, **kwds)
+        self.edge = edge
+
+    def __setattr__(self, name, value):
+        if name == 'name' and value and value.startswith('{'):
+            self.edge = int(value.strip('{}'))
+        else:
+            super().__setattr__(name, value)
diff --git a/bin/sv_table.R b/bin/sv_table.R
index 3bdc9bb..25bec2f 100644
--- a/bin/sv_table.R
+++ b/bin/sv_table.R
@@ -4,7 +4,7 @@ suppressPackageStartupMessages(library(argparse, quietly = TRUE))
 suppressPackageStartupMessages(library(dplyr, quietly = TRUE))
 suppressPackageStartupMessages(library(tidyr, quietly = TRUE))
 
-options(error=recover)
+options(error=NULL)
 options(width=200)
 
 concat <- function(...){
diff --git a/bin/unmerge.py b/bin/unmerge.py
new file mode 100755
index 0000000..e7b55da
--- /dev/null
+++ b/bin/unmerge.py
@@ -0,0 +1,70 @@
+#!/usr/bin/env python3
+import argparse
+import bz2
+import csv
+import io
+import jplace
+import json
+import sys
+
+from Bio import SeqIO
+
+
+def get_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('merged', type=argparse.FileType('r'))
+    parser.add_argument('--blast-details')
+    parser.add_argument('--jplace', type=argparse.FileType('r'))
+    parser.add_argument('--limit', type=int)
+    parser.add_argument('--name')
+    parser.add_argument('--seqs', type=argparse.FileType('r'))
+    parser.add_argument('--sort-by', default='likelihood')
+    parser.add_argument('--specimen')
+    parser.add_argument('--stockholm', type=argparse.FileType('r'))
+    parser.add_argument(
+        '--out', default=sys.stdout, type=argparse.FileType('w'))
+    return parser.parse_args()
+
+
+def main():
+    args = get_args()
+    names = {}  # {name: confidence}
+    if args.blast_details:
+        details = csv.DictReader(bz2.open(args.blast_details, 'rt'))
+        details = (d for d in details if d['specimen'] == args.specimen)
+        details = (d for d in details if d['pident'])
+        details = (d for d in details if d['assignment_id'])
+        details = sorted(
+            details, key=lambda x: float(x['pident']), reverse=True)
+        for d in details:
+            names[d['sseqid']] = float('-inf')
+    if args.name:
+        names[args.name] = 0
+    if args.stockholm:
+        for a in SeqIO.parse(args.stockholm, 'stockholm'):
+            names[a.name] = float('-inf')
+    if args.jplace:
+        jp = json.load(args.jplace)
+        tree = next(jplace.JParser(io.StringIO(jp['tree'])).parse())
+        pls = {}
+        if jp['placements']:
+            pls = jp['placements'][0]['p']
+            pls = [{f: p[i] for i, f in enumerate(jp['fields'])} for p in pls]
+            pls = {p['edge_num']: p for p in pls}
+            clades = (c for c in tree.find_clades() if c.edge in pls)
+            for c in clades:
+                if c.is_terminal():
+                    names[c.name] = pls[c.edge][args.sort_by]
+    seqs = SeqIO.parse(args.merged, 'fasta')
+    seqs = [s for s in seqs if s.name in names]
+    if not seqs and args.seqs:
+        seqs = SeqIO.parse(args.seqs, 'fasta')
+        seqs = [s for s in seqs if s.name in names]
+    seqs = sorted(seqs, key=lambda s: names[s.name], reverse=True)
+    if args.limit:
+        seqs = seqs[:args.limit]
+    SeqIO.write(seqs, args.out, 'fasta')
+
+
+if __name__ == '__main__':
+    sys.exit(main())
diff --git a/requirements.txt b/requirements.txt
index 3312205..2741e1a 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,4 +1,6 @@
-scons>=3.0.1
+biopython
 bioscons>=0.9.0
-
-
+csvkit
+fastalite
+pandas
+scons>=3.0.1