diff --git a/qiita_db/metadata_template/prep_template.py b/qiita_db/metadata_template/prep_template.py
index f69f162f1..19cf5dc5c 100644
--- a/qiita_db/metadata_template/prep_template.py
+++ b/qiita_db/metadata_template/prep_template.py
@@ -790,20 +790,22 @@ def _get_node_info(workflow, node):
                 ccmd.name, ccmd.merging_scheme, parent_cmd_name,
                 parent_merging_scheme, cparams, [], pparams)
 
-        def _get_predecessors(workflow, node):
-            # recursive method to get predecessors of a given node
-            pred = []
+        def _get_previous_predecessors(workflow, node):
+            parents, cxns = [], []
             for pnode in workflow.graph.predecessors(node):
-                pred = _get_predecessors(workflow, pnode)
-                cxns = {x[0]: x[2]
-                        for x in workflow.graph.get_edge_data(
-                            pnode, node)['connections'].connections}
-                data = [pnode, node, cxns]
-                if pred is None:
-                    pred = [data]
-                else:
-                    pred.append(data)
-                return pred
+                parents.append(pnode)
+                cxns.append(
+                    {x[0]: x[1] for x in workflow.graph.get_edge_data(
+                        pnode, node)['connections'].connections})
+            return [parents, node, cxns]
+
+        def _get_predecessors(workflow, node):
+            data = [_get_previous_predecessors(workflow, node)]
+            for pnode in data[0][0]:
+                d = _get_previous_predecessors(workflow, pnode)
+                if d[0]:
+                    data.append(d)
+            return data
 
         # Note: we are going to use the final BIOMs to figure out which
         #       processing is missing from the back/end to the front, as this
@@ -913,24 +915,30 @@ def _get_predecessors(workflow, node):
             previous_jobs = dict()
             for ma, node in wk_data.items():
                 predecessors = _get_predecessors(wk, node)
-                predecessors.reverse()
                 cmds_to_create = []
                 init_artifacts = None
-                for i, (pnode, cnode, cxns) in enumerate(predecessors):
+                for i, (pnodes, cnode, cxns) in enumerate(predecessors):
                     cdp = cnode.default_parameter
                     cdp_cmd = cdp.command
                     params = cdp.values.copy()
 
-                    icxns = {y: x for x, y in cxns.items()}
-                    reqp = {x: icxns[y[1][0]]
+                    # note that this assume that commands cannot have
+                    # multiple inputs with the same name, which should
+                    # be a safe assumption
+                    for pnode, cxn in zip(pnodes, cxns):
+                        info = _get_node_info(wk, pnode)
+                        if init_artifacts is None:
+                            init_artifacts = dict()
+                        for k, v in merging_schemes[info].items():
+                            if k in cxn:
+                                k = cxn[k]
+                            init_artifacts[k] = v
+
+                    reqp = {x: y[1][0]
                             for x, y in cdp_cmd.required_parameters.items()}
+
                     cmds_to_create.append([cdp_cmd, params, reqp])
 
-                    info = _get_node_info(wk, pnode)
-                    if info in merging_schemes:
-                        if set(merging_schemes[info]) >= set(cxns):
-                            init_artifacts = merging_schemes[info]
-                            break
                 if init_artifacts is None:
                     pdp = pnode.default_parameter
                     pdp_cmd = pdp.command
@@ -965,15 +973,15 @@ def _get_predecessors(workflow, node):
                         previous_job = current_job
                     if previous_job is None:
                         req_params = dict()
-                        for iname, dname in rp.items():
-                            if dname not in init_artifacts:
-                                msg = (f'Missing Artifact type: "{dname}" in '
+                        for aname, aid in rp.items():
+                            if aname not in init_artifacts:
+                                msg = (f'Missing Artifact type: "{aname}" in '
                                        'this preparation; this might be due '
                                        'to missing steps or not having the '
                                        'correct raw data.')
                                 # raises option c.
                                 raise ValueError(msg)
-                            req_params[iname] = init_artifacts[dname]
+                            req_params[aname] = init_artifacts[aname]
                     else:
                         req_params = dict()
                         connections = dict()