BODHI: Bharat Ontology for Disease & Healthcare Informatics

Open clinical knowledge graphs for grounding healthcare AI

SNOMED-linked knowledge graphs built to ground healthcare AI in verified clinical facts. Developed at Eka Care and battle-tested in production across symptom checking, differential diagnosis, and patient health profiling. Released openly under CC BY-NC 4.0.

→ Read the full writeup on the motivation, design, and use cases

Networks

Network	Focus	Nodes	Relationships
bodhi-s	Condition ↔ Symptom ↔ Speciality	4,855	13,204
bodhi-m	Concept ↔ Drug ↔ LabInvestigation	4,469	3,566

Downloads

Each network is available in five formats:

Format	File	Use case
Neo4j dump	`neo4j/*.dump`	Import into a local Neo4j instance
CSV	`csv/*.csv`	Flat-file processing, Pandas, SQL
JSONL	`jsonl/triples.jsonl`, `jsonl/nl_facts.jsonl`	LLM training, RAG pipelines
PyG	`pyg/*.pt`	PyTorch Geometric / GNN training
RDF/Turtle	`rdf/*.ttl`	Semantic web, SPARQL, ontology tools
Browser JSON	`browser_data_*.json`	Graph visualisation (nodes + edges in one file)

Loading the Neo4j dump

# Restore into a local Neo4j instance (Neo4j 5.x)
neo4j-admin database load --from-path=bodhi-s/neo4j/bodhi.dump --database=neo4j --overwrite-destination=true

Loading CSV files

import pandas as pd

nodes_condition  = pd.read_csv("bodhi-s/csv/nodes_condition.csv")
edges_present_in = pd.read_csv("bodhi-s/csv/edges_present_in.csv")

Loading PyG graphs

import torch
from torch_geometric.data import HeteroData

data = torch.load("bodhi-s/pyg/bodhi.pt")
print(data)

Querying RDF/Turtle

# Using Apache Jena's riot / arq
arq --data bodhi-s/rdf/bodhi.ttl --query your_query.sparql

bodhi-s — Condition-Symptom Network

Links conditions to symptoms, specialities, and inter-condition risk relationships. Symptoms are modelled as compound variants (e.g. Fever with chills, Fever for 3 days) with per-node triage levels and demographic likelihood scores.

Stats

Metric	Count
Condition nodes	779
Symptom nodes (variants)	4,037
Symptom root concepts (distinct SNOMED IDs)	590
Speciality nodes	39
Total relationships	13,204
Symptom → Condition edges (PRESENT_IN)	10,352
Condition → Speciality edges (TREATED_BY)	1,558
Condition → Condition edges (IS_INFLUENCED_BY)	1,020
Condition → Condition edges (RELATED_TO)	221
Condition → Condition edges (HAS_PREREQUISITE)	53

Condition types: Disorder 607 · Misc 84 · FamilyHistory 49 · Lifestyle 21 · Procedure 16 · Allergy 1 · Symptom 1

Avg symptoms per condition: 13.3

Condition triage: OPD Managed 367 (47%) · Worrisome 223 (29%) · Emergency 189 (24%)

Symptom triage: OPD Managed 2,244 (56%) · Worrisome 1,540 (38%) · Emergency 252 (6%)

Most cross-cutting symptoms: Fever (145 conditions) · Fatigue (126) · Headache (110) · Vomit (94) · Malaise (81)

Top specialities by condition volume: Internal Medicine (292) · General Physician (205) · Orthopedic (139) · Neurologist (83) · General Surgeon (81)

Schema

Node: `Condition`

A diagnosable medical condition, disorder, or clinical entity.

Property	Type	Values	Description
`snomed_id`	string	SNOMED CT ID	Globally unique clinical identifier
`name`	string	—	Clinical name of the condition
`concept_type`	enum	`Disorder` `Misc` `FamilyHistory` `Lifestyle` `Procedure` `Allergy` `Symptom`	Classification of the concept
`triage_level`	enum	`opd_managed` `worrisome` `emergency`	Clinical urgency
`type_condition`	enum	`acute` `chronic` `acute_that_may_turn_chronic` `chronic_with_acute_aggravation` `lifestyle` `medical_history` `Event` `Injury`	Temporal nature of condition
`overall_likelihood`	enum	`rare` `low` `medium` `high` `very_high`	Population prevalence signal
`likelihood_male`	float	0.0–1.0	Relative likelihood in males
`likelihood_female`	float	0.0–1.0	Relative likelihood in females
`likelihood_age_0_1`	float	0.0–1.0	Relative likelihood in age 0–1
`likelihood_age_1_5`	float	0.0–1.0	Relative likelihood in age 1–5
`likelihood_age_6_12`	float	0.0–1.0	Relative likelihood in age 6–12
`likelihood_age_13_18`	float	0.0–1.0	Relative likelihood in age 13–18
`likelihood_age_19_30`	float	0.0–1.0	Relative likelihood in age 19–30
`likelihood_age_30_45`	float	0.0–1.0	Relative likelihood in age 30–45
`likelihood_age_45_60`	float	0.0–1.0	Relative likelihood in age 45–60
`likelihood_age_60_plus`	float	0.0–1.0	Relative likelihood in age 60+

Node: `Symptom`

A clinical symptom or refinement thereof. Symptoms are structured with parent-child refinements (e.g. "Headache > Throbbing headache > Throbbing headache on right side").

Property	Type	Values	Description
`uuid`	string	UUID	Unique compound symptom identifier
`snomed_id`	string	SNOMED CT ID	SNOMED identifier for the symptom
`root_snomed_id`	string	SNOMED CT ID	Parent/root symptom SNOMED ID
`root_snomed_name`	string	—	Parent symptom name
`name`	string	—	Full compound symptom name
`triage_level`	enum	`opd_managed` `worrisome` `emergency`	Clinical urgency of this symptom
`relation1_type`	enum	`characteristic` `severity` `location` `laterality` `onset` `duration_since` `duration_lasts` `temporal_pattern` `pain_type` `radiating` `aggravated` `relieved`	Type of the first refinement axis
`child1_name`	string	—	Value of the first refinement
`grouping1_selection_type`	enum	`s` `m`	Single (`s`) or multi-select (`m`) for axis 1
`relation2_type`	enum	(same as relation1_type)	Type of the second refinement axis
`child2_name`	string	—	Value of the second refinement
`grouping2_selection_type`	enum	`s` `m`	Single or multi-select for axis 2
`relation3_type`	enum	(same as relation1_type)	Type of the third refinement axis
`child3_name`	string	—	Value of the third refinement
`grouping3_selection_type`	enum	`s` `m`	Single or multi-select for axis 3

Node: `Speciality`

A medical speciality or care discipline.

Property	Type	Description
`id`	string	Internal speciality identifier
`name`	string	Speciality name e.g. `Cardiologist`

Relationships

`(Symptom)-[:PRESENT_IN]->(Condition)`

Links a symptom to a condition it presents in. Encodes bidirectional likelihood.

Property	Values	Description
`likelihood_symptom_given_condition`	`zero` `rare` `low` `medium` `high` `very_high`	How commonly this symptom appears when the condition is present — P(symptom \| condition)
`likelihood_condition_given_symptom`	`zero` `rare` `low` `medium` `high` `very_high`	How predictive this symptom is of the condition — P(condition \| symptom)

`(Condition)-[:TREATED_BY]->(Speciality)`

Indicates which speciality manages a condition.

Property	Values	Description
`weight`	`rare` `low` `medium` `high` `very_high`	Strength of the referral association

`(Condition)-[:IS_INFLUENCED_BY]->(Condition)`

Condition A is clinically influenced by the presence of condition B (e.g. Diabetes IS_INFLUENCED_BY Obesity).

Property	Values	Description
`relation_strength`	`zero` `rare` `low` `medium` `high` `very_high`	Magnitude of influence
`relation_polarity`	`positive` `negative`	Positive = B increases risk of A; Negative = B decreases risk of A

`(Condition)-[:HAS_PREREQUISITE]->(Condition)`

Condition A requires condition B to be present (e.g. Diabetic nephropathy HAS_PREREQUISITE Diabetes mellitus).

Property	Values	Description
`relation_strength`	`medium` `high` `very_high`	How mandatory the prerequisite is
`relation_polarity`	`positive` `negative`	Direction of dependency

`(Symptom/Condition)-[:RELATED_TO]->(Symptom/Condition)`

Ontological relatedness — used for symptom deduplication and SNOMED hierarchy linkage.

Property	Values	Description
`relation_type`	`same_as` `similar_to` `parent_child`	Nature of the ontological relationship

bodhi-m — Concept-Drug-Lab Investigation Network

Maps SNOMED concepts (disorders, findings, procedures) to generic drugs and LOINC-coded lab investigations, organised in a three-level hierarchy (System → Group → Granular). Supports reverse inference from medications to conditions, and from lab results to health domains.

Stats

Metric	Count
Concept nodes	2,471
Drug nodes	1,186
LabInvestigation nodes	812
Total relationships	3,566
Concept → Concept edges (CHILD_OF)	1,768
Concept → Drug edges (TREATED_BY)	908
LabInvestigation → Concept edges (IMPACTS)	808
Concept → LabInvestigation edges (MONITORED_BY)	82

Concept hierarchy: System 14 → Group 250 → Granular 1,942 (+ 265 unmapped)

Hierarchy coverage: 1,540 / 1,942 granular linked to group · 228 / 250 groups linked to system

Schema

Node: `Concept`

A clinical concept — condition, disorder, finding, procedure, or lifestyle factor — organised in a three-level hierarchy: System → Group → Granular.

Property	Type	Values	Description
`snomed_id`	string	SNOMED CT ID	Primary unique identifier (open standard)
`name`	string	—	Clinical name
`display_name`	string	—	Consumer-friendly name e.g. `High Blood Pressure`
`level_concept`	enum	`system` `group` `granular`	Position in the clinical hierarchy
`type_concept`	enum	`Disorder` `Finding` `Procedure` `Lifestyle` `Allergy` `Situation`	Clinical category
`type_information`	enum	`SelfHistory` `FamilyHistory`	Whether this pertains to the patient themselves or family history
`active`	string	`1`	Whether this concept is active in the knowledge base

Hierarchy levels:

system — broad health domain e.g. Cardiovascular health, Endocrine health (14 nodes)
group — clinical cluster e.g. Diabetes mellitus, Coronary Artery Disease (250 nodes)
granular — specific diagnosable entity e.g. Diabetes mellitus type II (1,942 nodes)

Node: `Drug`

A generic drug formulation. Combination drugs are stored as a single node.

Property	Type	Description
`hash`	string	MD5 hash of the generic name — unique deduplication key
`name`	string	Generic drug name e.g. `metformin`, `atorvastatin + aspirin`
`therapeutic_class`	string	Comma-separated therapeutic class(es) e.g. `Anti Diabetic, Cardiovascular`

Node: `LabInvestigation`

A lab test or clinical measurement, identified by LOINC standard code.

Property	Type	Values	Description
`loinc_id`	string	LOINC ID	Globally unique lab test identifier (open standard)
`name`	string	—	Standard test name
`display_name`	string	—	Display/friendly name
`system_map`	string	—	Health domain e.g. `Renal health`, `Endocrine health`
`timespan_problem`	enum	`stat` `less_than_24_hr` `week_1` `month_1` `month_3` `month_6` `year_1` `lifetime`	How long this lab investigation result stays clinically relevant for a related condition
`impact_problem`	enum	`zero` `low` `medium` `high`	Clinical significance of this lab investigation in disease management

Relationships

`(Concept)-[:CHILD_OF]->(Concept)`

Encodes the three-level clinical hierarchy. Granular concepts point to their parent Group; Group concepts point to their parent System.

No properties.

`(LabInvestigation)-[:IMPACTS]->(Concept)`

A lab investigation broadly belongs to and impacts a health domain concept (always a system-level concept). Represents the primary health system this test monitors.

No properties.

`(Concept)-[:TREATED_BY]->(Drug)`

A clinical concept is treated by a generic drug. Encodes gender exclusivity signals for prescribing guidance.

Property	Values	Description
`therapeutic_class`	string	The drug class relevant for this specific indication
`exclusivity`	`zero` `low` `high`	How specific this drug is to this condition vs. used broadly
`exclusivity_male`	`low` `high`	Prescribing exclusivity signal for males
`exclusivity_female`	`low` `high`	Prescribing exclusivity signal for females

`(Concept)-[:MONITORED_BY]->(LabInvestigation)`

A condition is monitored or diagnostically associated with a specific lab test. Encodes the deduction power and directional threshold.

Property	Values	Description
`polarity`	`above` `below` `equal`	Which direction relative to the threshold is clinically significant
`category_threshold`	`normal` `borderline_low` `borderline_high` `high` `abnormal` `critically_high`	The result range that triggers this association
`vital_expiry_value`	`stat` `week_1` `month_1` `month_3` `month_6` `year_1`	The validity or the relevance of a test prescribed. Duration for which the vitals results are still considered relevant for a condition
`exclusivity`	`low` `high`	Deduction power — how strongly an abnormal result predicts this condition

Repository Structure

BODHI/
├── bodhi-s/
│   ├── csv/                        # Node and edge CSV files
│   │   ├── nodes_condition.csv
│   │   ├── nodes_symptom.csv
│   │   ├── nodes_speciality.csv
│   │   ├── edges_present_in.csv
│   │   ├── edges_treated_by.csv
│   │   ├── edges_is_influenced_by.csv
│   │   ├── edges_related_to.csv
│   │   └── edges_has_prerequisite.csv
│   ├── jsonl/
│   │   ├── triples.jsonl           # (subject, predicate, object) triples
│   │   └── nl_facts.jsonl          # Natural-language fact strings
│   ├── neo4j/
│   │   └── bodhi.dump              # Neo4j database dump
│   ├── pyg/
│   │   └── bodhi.pt                # PyTorch Geometric HeteroData object
│   ├── rdf/
│   │   └── bodhi.ttl               # RDF graph in Turtle format
│   └── browser_data_bodhi_s.json   # Combined nodes + edges for visualisation
└── bodhi-m/
    ├── csv/
    │   ├── nodes_concept.csv
    │   ├── nodes_drug.csv
    │   ├── nodes_lab_investigation.csv
    │   ├── edges_child_of.csv
    │   ├── edges_treated_by.csv
    │   ├── edges_impacts.csv
    │   └── edges_monitored_by.csv
    ├── jsonl/
    │   ├── triples.jsonl
    │   └── nl_facts.jsonl
    ├── neo4j/
    │   └── bodhi_m.dump
    ├── pyg/
    │   └── bodhi_m.pt
    ├── rdf/
    │   └── bodhi_m.ttl
    └── browser_data_bodhi_m.json

Standards Used

Standard	Used for
SNOMED CT	Condition and concept identifiers
LOINC	Lab investigation identifiers

License

This dataset is released under the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license.

You are free to share and adapt the data for non-commercial purposes, provided you give appropriate credit to Eka Care.

Name		Name	Last commit message	Last commit date
Latest commit History 1 Commit
bodhi-m		bodhi-m
bodhi-s		bodhi-s
LICENSE		LICENSE
README.md		README.md

Folders and files

Latest commit

History

Repository files navigation

BODHI: Bharat Ontology for Disease & Healthcare Informatics

Networks

Downloads

Loading the Neo4j dump

Loading CSV files

Loading PyG graphs

Querying RDF/Turtle

bodhi-s — Condition-Symptom Network

Stats

Schema

Node: Condition

Node: Symptom

Node: Speciality

Relationships

(Symptom)-[:PRESENT_IN]->(Condition)

(Condition)-[:TREATED_BY]->(Speciality)

(Condition)-[:IS_INFLUENCED_BY]->(Condition)

(Condition)-[:HAS_PREREQUISITE]->(Condition)

(Symptom/Condition)-[:RELATED_TO]->(Symptom/Condition)

bodhi-m — Concept-Drug-Lab Investigation Network

Stats

Schema

Node: Concept

Node: Drug

Node: LabInvestigation

Relationships

(Concept)-[:CHILD_OF]->(Concept)

(LabInvestigation)-[:IMPACTS]->(Concept)

(Concept)-[:TREATED_BY]->(Drug)

(Concept)-[:MONITORED_BY]->(LabInvestigation)

Repository Structure

Standards Used

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors 0

Languages

Node: `Condition`

Node: `Symptom`

Node: `Speciality`

`(Symptom)-[:PRESENT_IN]->(Condition)`

`(Condition)-[:TREATED_BY]->(Speciality)`

`(Condition)-[:IS_INFLUENCED_BY]->(Condition)`

`(Condition)-[:HAS_PREREQUISITE]->(Condition)`

`(Symptom/Condition)-[:RELATED_TO]->(Symptom/Condition)`

Node: `Concept`

Node: `Drug`

Node: `LabInvestigation`

`(Concept)-[:CHILD_OF]->(Concept)`

`(LabInvestigation)-[:IMPACTS]->(Concept)`

`(Concept)-[:TREATED_BY]->(Drug)`

`(Concept)-[:MONITORED_BY]->(LabInvestigation)`

Packages

Contributors