diff --git a/qiita_db/artifact.py b/qiita_db/artifact.py
index 2604ee6ef..e4ac92a34 100644
--- a/qiita_db/artifact.py
+++ b/qiita_db/artifact.py
@@ -1277,10 +1277,8 @@ def _add_edge(edges, src, dest):
             qdb.sql_connection.TRN.add(sql, [self.id])
             sql_edges = qdb.sql_connection.TRN.execute_fetchindex()
 
-            lineage = nx.DiGraph()
-            edges = set()
-            nodes = {}
-            if sql_edges:
+            # helper function to reduce code duplication
+            def _helper(sql_edges, edges, nodes):
                 for jid, pid, cid in sql_edges:
                     if jid not in nodes:
                         nodes[jid] = ('job',
@@ -1291,9 +1289,29 @@ def _add_edge(edges, src, dest):
                         nodes[cid] = ('artifact', qdb.artifact.Artifact(cid))
                     edges.add((nodes[pid], nodes[jid]))
                     edges.add((nodes[jid], nodes[cid]))
+
+            lineage = nx.DiGraph()
+            edges = set()
+            nodes = dict()
+            extra_edges = set()
+            extra_nodes = dict()
+            if sql_edges:
+                _helper(sql_edges, edges, nodes)
             else:
                 nodes[self.id] = ('artifact', self)
                 lineage.add_node(nodes[self.id])
+            # if this is an Analysis we need to check if there are extra
+            # edges/nodes as there is a chance that there are connecions
+            # between them
+            if self.analysis is not None:
+                roots = [a for a in self.analysis.artifacts
+                         if not a.parents and a != self]
+                for r in roots:
+                    # add the root to the options then their children
+                    extra_nodes[r.id] = ('artifact', r)
+                    qdb.sql_connection.TRN.add(sql, [r.id])
+                    sql_edges = qdb.sql_connection.TRN.execute_fetchindex()
+                    _helper(sql_edges, extra_edges, extra_nodes)
 
             # The code above returns all the jobs that have been successfully
             # executed. We need to add all the jobs that are in all the other
@@ -1329,8 +1347,10 @@ def _add_edge(edges, src, dest):
                             # need to check both the input_artifacts and the
                             # pending properties
                             for in_art in n_obj.input_artifacts:
-                                _add_edge(edges, nodes[in_art.id],
-                                          nodes[n_obj.id])
+                                iid = in_art.id
+                                if iid not in nodes and iid in extra_nodes:
+                                    nodes[iid] = extra_nodes[iid]
+                                _add_edge(edges, nodes[iid], nodes[n_obj.id])
 
                             pending = n_obj.pending
                             for pred_id in pending:
diff --git a/qiita_db/test/test_artifact.py b/qiita_db/test/test_artifact.py
index b9ba78149..5cd425e23 100644
--- a/qiita_db/test/test_artifact.py
+++ b/qiita_db/test/test_artifact.py
@@ -1406,6 +1406,42 @@ def test_has_human(self):
 
         self.assertTrue(artifact.has_human)
 
+    def test_descendants_with_jobs(self):
+        # let's tests that we can connect two artifacts with different root
+        # in the same analysis
+        # 1. make sure there are 3 nodes
+        a = qdb.artifact.Artifact(8)
+        self.assertEqual(len(a.descendants_with_jobs.nodes), 3)
+        self.assertEqual(len(a.analysis.artifacts), 2)
+        # 2. add a new root and make sure we see it
+        c = qdb.artifact.Artifact.create(
+            self.filepaths_root, "BIOM", analysis=a.analysis,
+            data_type="16S")
+        self.assertEqual(len(a.analysis.artifacts), 3)
+        # 3. add jobs conencting the new artifact to the other root
+        #    a -> job -> b
+        #    c
+        #    job1 connects b & c
+        #    job2 connects a & c
+        cmd = qdb.software.Command.create(
+            qdb.software.Software(1),
+            "CommandWithMultipleInputs", "", {
+                'input_b': ['artifact:["BIOM"]', None],
+                'input_c': ['artifact:["BIOM"]', None]}, {'out': 'BIOM'})
+        params = qdb.software.Parameters.load(
+            cmd, values_dict={'input_b': a.children[0].id, 'input_c': c.id})
+        job1 = qdb.processing_job.ProcessingJob.create(
+            qdb.user.User('test@foo.bar'), params)
+        params = qdb.software.Parameters.load(
+            cmd, values_dict={'input_b': a.id, 'input_c': c.id})
+        job2 = qdb.processing_job.ProcessingJob.create(
+            qdb.user.User('test@foo.bar'), params)
+
+        jobs = [j[1] for e in a.descendants_with_jobs.edges
+                for j in e if j[0] == 'job']
+        self.assertIn(job1, jobs)
+        self.assertIn(job2, jobs)
+
 
 @qiita_test_checker()
 class ArtifactArchiveTests(TestCase):
diff --git a/qiita_pet/handlers/analysis_handlers/base_handlers.py b/qiita_pet/handlers/analysis_handlers/base_handlers.py
index bc3de16d0..bd9c208e2 100644
--- a/qiita_pet/handlers/analysis_handlers/base_handlers.py
+++ b/qiita_pet/handlers/analysis_handlers/base_handlers.py
@@ -196,7 +196,9 @@ def analyisis_graph_handler_get_request(analysis_id, user):
                 # This should never happen, but worth having a useful message
                 raise ValueError('More than one workflow in a single analysis')
 
-    return {'edges': edges, 'nodes': nodes, 'workflow': wf_id,
+    # the list(set()) is to remove any duplicated nodes
+    return {'edges': list(set(edges)), 'nodes': list(set(nodes)),
+            'workflow': wf_id,
             'artifacts_being_deleted': artifacts_being_deleted}