fix(tier0): Fix Pinnedness Algorithm

[everaldorodrigo]

Summary

This PR refines the Pinnedness Algorithm used to select the starting node for multihop TRAPI→Dgraph queries. The fix ensures nodes with more-selective constraints (especially single IDs) are preferred.

Motivation

Previous pinnedness sometimes chose a less-constrained node when adjacency was dense.
Tests revealed cases where a node with an ID should have been the root but wasn’t.

The slow Dgraph query generated is below:

{
    q0_node_n2(func: eq(vF_category, "Disease")) @cascade(vF_id, ~vF_subject, ~vF_object) {
        expand(vF_Node)

        # Path 1: Disease (Subject) -> Phenotype (MONDO:0011705)
        out_edges_e1: ~vF_subject @filter(eq(vF_predicate_ancestors, "has_phenotype")) @cascade(vF_predicate, vF_object) {
            expand(vF_Edge) { vF_sources expand(vF_Source) }
            node_n0: vF_object @filter(eq(vF_id, "MONDO:0011705")) @cascade(vF_id) {
                expand(vF_Node)
            }
        }

        # Path 2: Chemical (Subject) -> Treatment (Edge) -> Disease (Object)
        in_edges_e0: ~vF_object @filter(eq(vF_predicate_ancestors, "treats_or_applied_or_studied_to_treat")) @cascade(vF_predicate, vF_subject) {
            expand(vF_Edge) { vF_sources expand(vF_Source) }
            node_n1: vF_subject @filter(eq(vF_category, "ChemicalEntity")) @cascade(vF_id) {
                expand(vF_Node)
            }
        }

        # Path 3: Phenotype (Subject) -> Phenotype (Edge) -> Disease (Object)
        in_edges_e2: ~vF_object @filter(eq(vF_predicate_ancestors, "has_phenotype")) @cascade(vF_predicate, vF_subject) {
            expand(vF_Edge) { vF_sources expand(vF_Source) }
            node_n0: vF_subject @filter(eq(vF_id, "MONDO:0011705")) @cascade(vF_id) {
                expand(vF_Node)
            }
        }
    }
}

The fixed Dgraph query generated is below:

{
    q0_node_n0(func: eq(vF_id, "MONDO:0011705")) @cascade(vF_id, ~vF_subject, ~vF_object) {
        expand(vF_Node)

        # Path 1: MONDO:0011705 is the OBJECT of 'has_phenotype' (i.e., Disease has MONDO:0011705)
        out_edges_e2: ~vF_subject @filter(eq(vF_predicate_ancestors, "has_phenotype")) @cascade(vF_predicate, vF_object) {
            expand(vF_Edge) { vF_sources expand(vF_Source) }
            node_n2: vF_object @filter(eq(vF_category, "Disease")) @cascade(vF_id, ~vF_object) {
                expand(vF_Node)
                # Find Chemical Treatments for this Disease
                in_edges_e0: ~vF_object @filter(eq(vF_predicate_ancestors, "treats_or_applied_or_studied_to_treat")) @cascade(vF_predicate, vF_subject) {
                    expand(vF_Edge) { vF_sources expand(vF_Source) }
                    node_n1: vF_subject @filter(eq(vF_category, "ChemicalEntity")) @cascade(vF_id) {
                        expand(vF_Node)
                    }
                }
            }
        }

        # Path 2: MONDO:0011705 is the SUBJECT of 'has_phenotype' (i.e., MONDO:0011705 has Disease as a phenotype)
        in_edges_e1: ~vF_object @filter(eq(vF_predicate_ancestors, "has_phenotype")) @cascade(vF_predicate, vF_subject) {
            expand(vF_Edge) { vF_sources expand(vF_Source) }
            node_n2: vF_subject @filter(eq(vF_category, "Disease")) @cascade(vF_id, ~vF_object) {
                expand(vF_Node)
                # Find Chemical Treatments for this Disease
                in_edges_e0: ~vF_object @filter(eq(vF_predicate_ancestors, "treats_or_applied_or_studied_to_treat")) @cascade(vF_predicate, vF_subject) {
                    expand(vF_Edge) { vF_sources expand(vF_Source) }
                    node_n1: vF_subject @filter(eq(vF_category, "ChemicalEntity")) @cascade(vF_id) {
                        expand(vF_Node)
                    }
                }
            }
        }
    }
}

[sentry]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
...c/retriever/data_tiers/tier_0/dgraph/transpiler.py	79.62% <100.00%> (+0.80%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.