Update data/embeddings/opa2vec_goa_human/README.md
Browse files
data/embeddings/opa2vec_goa_human/README.md
CHANGED
|
@@ -1,9 +1,7 @@
|
|
| 1 |
# Embedding name: OPA2Vec-GOA-human
|
| 2 |
|
| 3 |
- Modality: Gene Ontology / ontology-annotation semantic embedding
|
| 4 |
-
- Algorithm: OPA2Vec-derived GOA-human semantic embedding generated locally from GO and GOA human inputs
|
| 5 |
- Embedding source URL: https://github.com/bio-ontology-research-group/opa2vec
|
| 6 |
-
- Original model/code URL: https://github.com/bio-ontology-research-group/opa2vec
|
| 7 |
|
| 8 |
## Public files
|
| 9 |
|
|
@@ -43,7 +41,7 @@
|
|
| 43 |
|
| 44 |
This embedding represents genes using an OPA2Vec-derived semantic representation built from Gene Ontology structure, GO metadata, and GOA human UniProt-GO annotations.
|
| 45 |
|
| 46 |
-
This should be interpreted as a GO/GOA semantic gene embedding. It captures ontology and annotation context
|
| 47 |
|
| 48 |
### Provenance fields
|
| 49 |
|
|
@@ -51,8 +49,6 @@ This should be interpreted as a GO/GOA semantic gene embedding. It captures onto
|
|
| 51 |
original_model_or_method: OPA2Vec-derived GOA-human
|
| 52 |
input_data_type: GO basic ontology, GOA human annotations, and locally generated ontology/annotation corpus
|
| 53 |
raw_source: ~/data/raw_embeddings/opa2vec_original_try/go_human_inputs/
|
| 54 |
-
processing_script: ~/scripts/process_embeddings/process_opa2vec_goa_human_v1_1.py
|
| 55 |
-
mapping_or_qc_report: ~/reports/mapping_reports/opa2vec_goa_human_v1_1_mapping_report.txt
|
| 56 |
source_download_page: https://github.com/bio-ontology-research-group/opa2vec
|
| 57 |
source_code_page: https://github.com/bio-ontology-research-group/opa2vec
|
| 58 |
source_identifier_type: UniProt accession from GOA human annotation-derived entity vectors
|
|
@@ -65,11 +61,9 @@ This should be interpreted as a GO/GOA semantic gene embedding. It captures onto
|
|
| 65 |
|
| 66 |
### Source-processing decisions
|
| 67 |
|
| 68 |
-
The final OPA2Vec matrix is
|
| 69 |
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
This decision is important for reproducibility: someone repeating the project should not expect to run the old OPA2Vec repository unchanged and obtain this exact matrix. They should instead rerun the recorded local processing workflow and inputs.
|
| 73 |
|
| 74 |
### Mapping and duplicate-handling decisions
|
| 75 |
|
|
@@ -94,9 +88,7 @@ For this embedding:
|
|
| 94 |
|
| 95 |
### Important implementation decisions
|
| 96 |
|
| 97 |
-
|
| 98 |
-
|
| 99 |
-
For OPA2Vec, the matrix is explicitly described as locally generated and OPA2Vec-derived because the final reproducible workflow reconstructed the ontology/annotation embedding process in Python 3 rather than using the old official implementation unchanged. This means the retained vectors should be reproduced from the recorded local workflow and inputs, not by expecting the old repository to run unchanged.
|
| 100 |
|
| 101 |
### Summary of final mapping:
|
| 102 |
|
|
|
|
| 1 |
# Embedding name: OPA2Vec-GOA-human
|
| 2 |
|
| 3 |
- Modality: Gene Ontology / ontology-annotation semantic embedding
|
|
|
|
| 4 |
- Embedding source URL: https://github.com/bio-ontology-research-group/opa2vec
|
|
|
|
| 5 |
|
| 6 |
## Public files
|
| 7 |
|
|
|
|
| 41 |
|
| 42 |
This embedding represents genes using an OPA2Vec-derived semantic representation built from Gene Ontology structure, GO metadata, and GOA human UniProt-GO annotations.
|
| 43 |
|
| 44 |
+
This should be interpreted as a GO/GOA semantic gene embedding. It captures ontology and annotation context.
|
| 45 |
|
| 46 |
### Provenance fields
|
| 47 |
|
|
|
|
| 49 |
original_model_or_method: OPA2Vec-derived GOA-human
|
| 50 |
input_data_type: GO basic ontology, GOA human annotations, and locally generated ontology/annotation corpus
|
| 51 |
raw_source: ~/data/raw_embeddings/opa2vec_original_try/go_human_inputs/
|
|
|
|
|
|
|
| 52 |
source_download_page: https://github.com/bio-ontology-research-group/opa2vec
|
| 53 |
source_code_page: https://github.com/bio-ontology-research-group/opa2vec
|
| 54 |
source_identifier_type: UniProt accession from GOA human annotation-derived entity vectors
|
|
|
|
| 61 |
|
| 62 |
### Source-processing decisions
|
| 63 |
|
| 64 |
+
The final OPA2Vec matrix is locally generated.
|
| 65 |
|
| 66 |
+
The official OPA2Vec route was investigated. The original implementation depended on older tooling and sidecar files that were not practical in the current workstation environment. Rather than keeping a failed or non-reproducible run, the final retained workflow was a reproducible Python 3 reconstruction: GO basic ontology and GOA human annotations were parsed locally, ontology/annotation context was converted into a text-like corpus, and Word2Vec-style training was used to obtain 200-dimensional vectors.
|
|
|
|
|
|
|
| 67 |
|
| 68 |
### Mapping and duplicate-handling decisions
|
| 69 |
|
|
|
|
| 88 |
|
| 89 |
### Important implementation decisions
|
| 90 |
|
| 91 |
+
The central requirement was that every retained row must align to one gene in master_gene_table.csv and that the matrix row order must match genes.tsv exactly.
|
|
|
|
|
|
|
| 92 |
|
| 93 |
### Summary of final mapping:
|
| 94 |
|