> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pav.bio/llms.txt
> Use this file to discover all available pages before exploring further.

# Build a competitive landscape

> Collect every program against a target, count it by phase, and pull its trials and nearest competitors.

This guide builds a landscape of KRAS-directed programs: who is developing
them, how far along each is, and which trials back them. The same pattern
works for an indication or a modality.

## Prerequisites

* A Pav API key in `PAV_API_KEY`. See [Quickstart](/quickstart).
* Python 3.9+ and `pip install requests`.

## Approach

1. **Search by meaning.** `q` finds programs whose text describes the
   mechanism, including ones with no structured `target`.
2. **Add exact target matches.** The `target` filter catches programs that
   state the target but word their description differently.
3. **Merge by program `id`, then filter.** Relevance search returns weaker
   matches further down the list; keep only programs that name the target.
4. **Rank by phase** and read the linked trials.

Running both searches matters: only about a quarter of programs publish a
target (see [Coverage and sources](/concepts/coverage-and-sources)).

## Script

```python landscape.py theme={null}
import collections
import os

import requests

API = "https://api.pav.bio"
HEADERS = {"Authorization": f"Bearer {os.environ['PAV_API_KEY']}"}


def get(path, **params):
    resp = requests.get(f"{API}{path}", headers=HEADERS, params=params, timeout=60)
    resp.raise_for_status()
    return resp.json()


# 1. Search by meaning, then add exact target matches.
found = {}
for query in ["KRAS G12C inhibitor", "pan-KRAS inhibitor"]:
    for p in get("/v1/programs", q=query, limit=200)["data"]:
        found[p["id"]] = p
for target in ["KRAS G12C", "KRAS"]:
    for p in get("/v1/programs", target=target, limit=500)["data"]:
        found[p["id"]] = p


# 2. Keep programs that name KRAS in their target, mechanism or drug.
def mentions_kras(p):
    text = " ".join([p.get("target", ""), p.get("mechanism_of_action", ""), p["drug"]])
    return "KRAS" in text.upper()


landscape = [p for p in found.values() if mentions_kras(p)]

# 3. Count by phase and list the most advanced programs.
RANK = ["Approved", "Registration", "Filed", "Phase 3", "Phase 2/3", "Phase 2",
        "Phase 1/2", "Phase 1", "Preclinical", "Unknown"]
by_phase = collections.Counter(p.get("phase_norm", "Unknown") for p in landscape)
print(f"{len(landscape)} KRAS programs across {len({p['company_id'] for p in landscape})} companies")
for phase in RANK:
    if by_phase[phase]:
        print(f"  {phase:<12} {by_phase[phase]}")

landscape.sort(key=lambda p: RANK.index(p.get("phase_norm", "Unknown")))
print()
for p in landscape[:8]:
    trials = len(p.get("clinical_trials", []))
    print(f"{p.get('phase_norm', 'Unknown'):<10} | {p['company']} | {p['drug']} | {p.get('indication', '')[:40]} | trials={trials}")
```

```text Example output (September 2026) theme={null}
50 KRAS programs across 27 companies
  Approved     2
  Registration 1
  Phase 3      16
  Phase 2      6
  Phase 1/2    9
  Phase 1      12
  Preclinical  4

Approved   | Shanghai Allist Pharmaceuticals | Glecirasib | 2nd Line Treatment for NSCLC | trials=10
Approved   | Innovent Biologics | DUPERT (Fulzerasib) |  | trials=15
Registration | Royalty Pharma | Rasonque | RAS mutant pancreatic cancer | trials=13
Phase 3    | Merck | calderasib | Colorectal cancer | trials=1
Phase 3    | Merck | calderasib | Non-small cell lung cancer | trials=5
Phase 3    | Bristol Myers Squibb | KRAZATI® (adagrasib) | 1L Non-Small Cell Lung Cancer PD-L1≥50% | trials=1
Phase 3    | Roche | divarasib (KRAS G12C) | 1L non small cell lung cancer (1L NSCLC) | trials=5
Phase 3    | Roche | divarasib (KRAS G12C) | 2L non small cell lung cancer (2L NSCLC) | trials=5
```

A program is one drug in one indication, so a drug developed in several
indications appears once per indication. Count distinct `drug` per company for
an asset-level view.

## Extend it

* **Nearest competitors of one asset.** `GET /v1/programs/{id}/similar-programs`
  returns the programs most similar to a given program. Run it on the leading
  assets to catch programs your query terms missed.
* **Trial detail.** Each program lists `clinical_trials[]` with `nct_id`,
  phase, status and sponsor. Fetch `GET /v1/trials/{nct_id}` for enrollment,
  dates and primary outcomes.
* **By indication.** Replace the searches with `q="obesity"` plus
  `indication=Obesity`, and group by `indication_terms[].term_id` to merge
  phrasings such as "Obesity" and "Metabolic (Obesity)".
* **Deals and patents around the landscape.** Pass each company's
  `company_id` to [`/v1/deals`](/datasets/deals) and [`/v1/patents`](/datasets/patents).

With the [MCP server](/mcp-server), an agent can do the same from one prompt:
"Build a landscape of KRAS inhibitors by phase with their lead indications and trial counts."