Merge pull request #81 from monarch-initiative/fix-genesets

fixing gene sets and updating analysis
monarch-initiative · May 2, 2023 · df38985 · df38985
2 parents 81a243d + 0da860e
commit df38985
Show file tree

Hide file tree

Showing 55 changed files with 5,219 additions and 12,308 deletions.
diff --git a/Makefile b/Makefile
@@ -130,7 +130,7 @@ analysis/enrichment/%-results-$(N).yaml: tests/input/genesets/%.yaml
 	$(RUN) ontogpt -vv eval-enrichment -n $(N) -U $< -o $@.tmp && mv $@.tmp $@
 
 analysis/enrichment-summary.yaml:
-	cat analysis/enrichment/*-$(N)yaml > $@
+	cat analysis/enrichment/*-$(N).yaml > $@
 
 analysis/enrichment-summary-$(N).yaml:
 	cat analysis/enrichment/*-$(N).yaml > $@

diff --git a/notebooks/Enrichment-Results-Analysis.ipynb b/notebooks/Enrichment-Results-Analysis.ipynb
diff --git a/src/ontogpt/engines/enrichment.py b/src/ontogpt/engines/enrichment.py
@@ -5,7 +5,7 @@
 from dataclasses import dataclass, field
 from enum import Enum
 from pathlib import Path
-from typing import Dict, List, Optional, Tuple
+from typing import Dict, List, Optional, Tuple, Union
 
 from jinja2 import Template
 from oaklib import BasicOntologyInterface, get_adapter
@@ -120,9 +120,9 @@ def summarize(
                 raise NotImplementedError
         if not gene_set.gene_ids and not gene_set.gene_symbols:
             raise ValueError(f"Gene set {gene_set.name} has no gene symbols or ids")
-        if gene_set.gene_ids and not gene_set.gene_symbols:
-            adapter = list(self.label_resolvers.values())[0]
-            gene_set.gene_symbols = [adapter.label(x.lower()) for x in gene_set.gene_ids]
+        # if gene_set.gene_ids and not gene_set.gene_symbols:
+        #    adapter = list(self.label_resolvers.values())[0]
+        #    gene_set.gene_symbols = [adapter.label(x.lower()) for x in gene_set.gene_ids]
         if not gene_set.gene_ids or normalize:
             gene_set.gene_ids = list(self.map_labels(gene_set.gene_symbols, strict=strict))
             logger.info(f"gene ids: {gene_set.gene_ids}")
@@ -152,7 +152,10 @@ def summarize(
         if not prompt_template:
             prompt_template = str(f"{DEFAULT_ENRICHMENT_PROMPT}.jinja2")
         prompt, tf = self._prompt_from_template(
-            gene_tuples, template=prompt_template, annotations=annotations
+            gene_tuples,
+            template=prompt_template,
+            annotations=annotations,
+            taxon=gene_set.taxon,
         )
         response_text = self.client.complete(prompt, max_tokens=self.completion_length)
         response_token_length = len(self.encoding.encode(response_text))
@@ -173,8 +176,9 @@ def summarize(
     def _prompt_from_template(
         self,
         genes: List[GENE_TUPLE],
-        template: str,
+        template: Union[str, Path, Template],
         truncation_factor=1.0,
+        taxon: str = None,
         annotations=True,
     ) -> Tuple[str, float]:
         if isinstance(template, Path):
@@ -195,22 +199,26 @@ def _prompt_from_template(
         prompt = template.render(
             gene_descriptions=gd_tuples,
             annotations=annotations,
+            taxon=taxon,
             SUMMARY_KEYWORD=SUMMARY_KEYWORD,
             MECHANISM_KEYWORD=MECHANISM_KEYWORD,
             ENRICHED_TERMS_KEYWORD=ENRICHED_TERMS_KEYWORD,
         )
         logging.debug(f"Prompt from template: {prompt}")
         logging.info(f"Prompt [{truncation_factor}] Length: {len(prompt)}")
         # https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them
-        prompt_length = len(self.encoding.encode(prompt))
-        logging.info(f"Prompt [{truncation_factor}] Tokens: {prompt_length} Strlen={len(prompt)}")
+        prompt_length = len(self.encoding.encode(prompt)) + 10
         max_len = 4097 - self.completion_length
+        logging.info(
+            f"Prompt [{truncation_factor}] Toks: {prompt_length} / {max_len} Str={len(prompt)}"
+        )
         if prompt_length > max_len:  # TODO: check this
             logging.warning(f"Prompt is too long; toks: {prompt_length} len: {len(prompt)}")
             return self._prompt_from_template(
                 genes,
                 template,
                 truncation_factor=truncation_factor * 0.8,
+                taxon=taxon,
                 annotations=annotations,
             )
         return prompt, truncation_factor

diff --git a/src/ontogpt/prompts/enrichment/gene_set_summarization.jinja2 b/src/ontogpt/prompts/enrichment/gene_set_summarization.jinja2
@@ -1,8 +1,4 @@
-{% if annotations %}
-I will give you a list of genes together with descriptions of their functions.
-{% else %}
-I will give you a list of genes.
-{% endif %}
+I will give you a list of {{ taxon }} genes {% if annotations %} together with descriptions of their functions{% endif %}.
 Perform a term enrichment test on these genes.
 i.e. tell me what the commonalities are in their function.
 Make use of classification hierarchies when you do this.

diff --git a/src/ontogpt/utils/gene_set_utils.py b/src/ontogpt/utils/gene_set_utils.py
@@ -26,7 +26,10 @@ class GeneSet(BaseModel):
     gene_symbols: Optional[List[str]] = None
     gene_ids: Optional[List[str]] = None
     taxon: str = "human"
+    taxon_id: Optional[str] = None
     description: Optional[str] = None
+    source: Optional[str] = None
+    source_url: Optional[str] = None
     target_term_ids: Optional[List[str]] = None
 
 

diff --git a/tests/input/genesets/HALLMARK_ADIPOGENESIS.yaml b/tests/input/genesets/HALLMARK_ADIPOGENESIS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ADIPOGENESIS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCA1
  - ABCB8
  - ACAA2

diff --git a/tests/input/genesets/HALLMARK_ALLOGRAFT_REJECTION.yaml b/tests/input/genesets/HALLMARK_ALLOGRAFT_REJECTION.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ALLOGRAFT_REJECTION
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - AARS1
  - ABCE1
  - ABI1

diff --git a/tests/input/genesets/HALLMARK_ANDROGEN_RESPONSE.yaml b/tests/input/genesets/HALLMARK_ANDROGEN_RESPONSE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ANDROGEN_RESPONSE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCC4
  - ABHD2
  - ACSL3

diff --git a/tests/input/genesets/HALLMARK_ANGIOGENESIS.yaml b/tests/input/genesets/HALLMARK_ANGIOGENESIS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ANGIOGENESIS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - APOH
  - APP
  - CCND2

diff --git a/tests/input/genesets/HALLMARK_APICAL_JUNCTION.yaml b/tests/input/genesets/HALLMARK_APICAL_JUNCTION.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_APICAL_JUNCTION
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ACTA1
  - ACTB
  - ACTC1

diff --git a/tests/input/genesets/HALLMARK_APICAL_SURFACE.yaml b/tests/input/genesets/HALLMARK_APICAL_SURFACE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_APICAL_SURFACE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ADAM10
  - ADIPOR2
  - AFAP1L2

diff --git a/tests/input/genesets/HALLMARK_APOPTOSIS.yaml b/tests/input/genesets/HALLMARK_APOPTOSIS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_APOPTOSIS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ADD1
  - AIFM3
  - ANKH

diff --git a/tests/input/genesets/HALLMARK_BILE_ACID_METABOLISM.yaml b/tests/input/genesets/HALLMARK_BILE_ACID_METABOLISM.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_BILE_ACID_METABOLISM
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCA1
  - ABCA2
  - ABCA3

diff --git a/tests/input/genesets/HALLMARK_CHOLESTEROL_HOMEOSTASIS.yaml b/tests/input/genesets/HALLMARK_CHOLESTEROL_HOMEOSTASIS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_CHOLESTEROL_HOMEOSTASIS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCA2
  - ACAT2
  - ACSS2

diff --git a/tests/input/genesets/HALLMARK_COAGULATION.yaml b/tests/input/genesets/HALLMARK_COAGULATION.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_COAGULATION
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - A2M
  - ACOX2
  - ADAM9

diff --git a/tests/input/genesets/HALLMARK_COMPLEMENT.yaml b/tests/input/genesets/HALLMARK_COMPLEMENT.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_COMPLEMENT
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ACTN2
  - ADAM9
  - ADRA2B

diff --git a/tests/input/genesets/HALLMARK_DNA_REPAIR.yaml b/tests/input/genesets/HALLMARK_DNA_REPAIR.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_DNA_REPAIR
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - AAAS
  - ADA
  - ADCY6

diff --git a/tests/input/genesets/HALLMARK_E2F_TARGETS.yaml b/tests/input/genesets/HALLMARK_E2F_TARGETS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_E2F_TARGETS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - AK2
  - ANP32E
  - ASF1A

diff --git a/tests/input/genesets/HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION.yaml b/tests/input/genesets/HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABI3BP
  - ACTA2
  - ADAM12

diff --git a/tests/input/genesets/HALLMARK_ESTROGEN_RESPONSE_EARLY.yaml b/tests/input/genesets/HALLMARK_ESTROGEN_RESPONSE_EARLY.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ESTROGEN_RESPONSE_EARLY
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABAT
  - ABCA3
  - ABHD2

diff --git a/tests/input/genesets/HALLMARK_ESTROGEN_RESPONSE_LATE.yaml b/tests/input/genesets/HALLMARK_ESTROGEN_RESPONSE_LATE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_ESTROGEN_RESPONSE_LATE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCA3
  - ABHD2
  - ACOX2

diff --git a/tests/input/genesets/HALLMARK_FATTY_ACID_METABOLISM.yaml b/tests/input/genesets/HALLMARK_FATTY_ACID_METABOLISM.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_FATTY_ACID_METABOLISM
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - AADAT
  - ACAA1
  - ACAA2

diff --git a/tests/input/genesets/HALLMARK_G2M_CHECKPOINT.yaml b/tests/input/genesets/HALLMARK_G2M_CHECKPOINT.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_G2M_CHECKPOINT
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABL1
  - AMD1
  - ARID4A

diff --git a/tests/input/genesets/HALLMARK_GLYCOLYSIS.yaml b/tests/input/genesets/HALLMARK_GLYCOLYSIS.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_GLYCOLYSIS
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCB6
  - ADORA2B
  - AGL

diff --git a/tests/input/genesets/HALLMARK_HEDGEHOG_SIGNALING.yaml b/tests/input/genesets/HALLMARK_HEDGEHOG_SIGNALING.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_HEDGEHOG_SIGNALING
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ACHE
  - ADGRG1
  - AMOT

diff --git a/tests/input/genesets/HALLMARK_HEME_METABOLISM.yaml b/tests/input/genesets/HALLMARK_HEME_METABOLISM.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_HEME_METABOLISM
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCB6
  - ABCG2
  - ACKR1

diff --git a/tests/input/genesets/HALLMARK_HYPOXIA.yaml b/tests/input/genesets/HALLMARK_HYPOXIA.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_HYPOXIA
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ACKR3
  - ADM
  - ADORA2B

diff --git a/tests/input/genesets/HALLMARK_IL2_STAT5_SIGNALING.yaml b/tests/input/genesets/HALLMARK_IL2_STAT5_SIGNALING.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_IL2_STAT5_SIGNALING
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCB1
  - ADAM19
  - AGER

diff --git a/tests/input/genesets/HALLMARK_IL6_JAK_STAT3_SIGNALING.yaml b/tests/input/genesets/HALLMARK_IL6_JAK_STAT3_SIGNALING.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_IL6_JAK_STAT3_SIGNALING
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - A2M
  - ACVR1B
  - ACVRL1

diff --git a/tests/input/genesets/HALLMARK_INFLAMMATORY_RESPONSE.yaml b/tests/input/genesets/HALLMARK_INFLAMMATORY_RESPONSE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_INFLAMMATORY_RESPONSE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCA1
  - ABI1
  - ACVR1B

diff --git a/tests/input/genesets/HALLMARK_INTERFERON_ALPHA_RESPONSE.yaml b/tests/input/genesets/HALLMARK_INTERFERON_ALPHA_RESPONSE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_INTERFERON_ALPHA_RESPONSE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ADAR
  - B2M
  - BATF2

diff --git a/tests/input/genesets/HALLMARK_INTERFERON_GAMMA_RESPONSE.yaml b/tests/input/genesets/HALLMARK_INTERFERON_GAMMA_RESPONSE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_INTERFERON_GAMMA_RESPONSE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ADAR
  - APOL6
  - ARID5B

diff --git a/tests/input/genesets/HALLMARK_KRAS_SIGNALING_DN.yaml b/tests/input/genesets/HALLMARK_KRAS_SIGNALING_DN.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_KRAS_SIGNALING_DN
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCB11
  - ABCG4
  - ACTC1

diff --git a/tests/input/genesets/HALLMARK_KRAS_SIGNALING_UP.yaml b/tests/input/genesets/HALLMARK_KRAS_SIGNALING_UP.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_KRAS_SIGNALING_UP
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCB1
  - ACE
  - ADAM17

diff --git a/tests/input/genesets/HALLMARK_MITOTIC_SPINDLE.yaml b/tests/input/genesets/HALLMARK_MITOTIC_SPINDLE.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_MITOTIC_SPINDLE
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABI1
  - ABL1
  - ABR

diff --git a/tests/input/genesets/HALLMARK_MTORC1_SIGNALING.yaml b/tests/input/genesets/HALLMARK_MTORC1_SIGNALING.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_MTORC1_SIGNALING
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCF2
  - ACACA
  - ACLY

diff --git a/tests/input/genesets/HALLMARK_MYC_TARGETS_V1.yaml b/tests/input/genesets/HALLMARK_MYC_TARGETS_V1.yaml
@@ -1,6 +1,6 @@
 name: HALLMARK_MYC_TARGETS_V1
-gene_symbols: []
-gene_ids:
+
+gene_symbols:
  - ABCE1
  - ACP1
  - AIMP2