---
title: "Predict IVIVE inputs and clearance for many structures"
method: POST
path: "/api/prediction/structures"
tags: ["prediction"]
---

# Predict IVIVE inputs and clearance for many structures

`POST /api/prediction/structures`

## Request

- Base URL: `https://api.ivive.app`
- URL: `https://api.ivive.app/api/prediction/structures`
- Auth: OAuth 2 (password, token URL `/api/auth/login`)

Up to 500 structures through the same models as a single prediction: descriptors, fu,p with its range, fu,inc, and the structure-only CL and Vss. A structure that cannot be parsed is reported in its own row and does not fail the request. No measured input is involved, so these are predictions throughout.

## Request body

- StructureBatchRequest: Structures to predict for. Validation is per row, so one bad SMILES keeps the rest.
  - `molecules` NamedStructure[], required
    - `name` string
    - `smiles` string, required
  - `assay_type` 'microsomes' | 'hepatocytes'
  - `concentration` number

## Response `200`

Successful Response

- StructureBatchResponse
  - `molecules` StructurePrediction[], required
    - `name` string, required
    - `smiles` string, required
    - `mw` number, nullable
    - `logp` number, nullable
    - `tpsa` number, nullable
    - `fu_p` RangedValue
      - `value` number, required
      - `low` number, required: Lower end of the range holding the measured value for half of held-out drugs.
      - `high` number, required
      - `coverage` number, required
    - `fu_inc` number, nullable
    - `cl` StructureCL: Human plasma CL predicted from structure alone: a second opinion, not an IVIVE input.
      - `value` number, required: Plasma clearance, mL/min/kg.
      - `low` number, required
      - `high` number, required
      - `coverage` number, required
      - `applicability` Applicability, required
        - `max_similarity` number, required: Tanimoto (Morgan r2) to the nearest fu,p training structure.
        - `status` 'in_domain' | 'borderline' | 'out_of_domain', required
    - `vdss` StructureVdss: Human Vss predicted from structure alone: a second opinion on the Schmitt 2008 result.
      - `value` number, required: Volume of distribution at steady state, L/kg.
      - `low` number, required
      - `high` number, required
      - `coverage` number, required
      - `applicability` Applicability, required
        - `max_similarity` number, required: Tanimoto (Morgan r2) to the nearest fu,p training structure.
        - `status` 'in_domain' | 'borderline' | 'out_of_domain', required
    - `error` string, nullable: Why this structure was skipped.
  - `n_predicted` integer, required
  - `n_skipped` integer, required

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/ivive/apis/ivive-app-api.md) · [All operations](https://skmtc.dev/ivive/apis/ivive-app-api/llms.txt) · [OpenAPI document](https://skmtc.dev/ivive/apis/ivive-app-api/revisions/a67615eaf584?raw)