GraphQL API Modernization

The key challenge was creating a schema that felt intuitive while mapping to complex underlying data:

• Clinical Trials: Unified model across ClinicalTrials.gov, EudraCT, and proprietary databases
• Companies: Merged data from SEC filings, LinkedIn, and Crunchbase
• Products: Drug pipeline data with regulatory milestone tracking

Our schema design principles:

1. Domain-driven types — Reflect pharmaceutical industry concepts, not database tables
2. Connections over IDs — Use GraphQL connections for relationships
3. Computed fields — Business logic like “trial success probability” computed on-demand
4. Nullable by default — Pharmaceutical data is often incomplete

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15

type ClinicalTrial {
  id: ID!
  nctId: String
  title: String!
  phase: TrialPhase!
  status: TrialStatus!
  sponsor: Company!
  products: [Product!]!
  milestones: [Milestone!]!
  
  # Computed fields
  successProbability: Float
  estimatedCompletionDate: Date
  competingTrials: [ClinicalTrial!]!
}

Real-time updates were critical for users monitoring clinical trial progress.

Architecture for Scale:

• PostgreSQL NOTIFY/LISTEN for change events
• Hasura converts changes to subscription payloads
• Client-side Apollo cache updates automatically

Optimization Strategies:

• Subscription multiplexing — single connection handles multiple subscriptions
• Smart batching — aggregate rapid changes before pushing
• Connection pooling — limit simultaneous WebSocket connections

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

const TRIAL_UPDATES = gql`
  subscription TrialUpdates($companyId: uuid!) {
    clinical_trials(
      where: { sponsor_id: { _eq: $companyId } }
      order_by: { updated_at: desc }
      limit: 50
    ) {
      id
      status
      phase
      updated_at
    }
  }
`;