danieltslz commited on
Commit
ff1f1a4
·
verified ·
1 Parent(s): 552f1c8

Update data/embeddings/frogs_archs4_256/README.md

Browse files
data/embeddings/frogs_archs4_256/README.md CHANGED
@@ -1,9 +1,7 @@
1
  # Embedding name: FRoGS-ARCHS4-256
2
 
3
  - Modality: Co-expression / ARCHS4 functional expression embedding
4
- - Algorithm: Official FRoGS ARCHS4 gene vectors standardized by strict Entrez mapping
5
  - Embedding source URL: https://github.com/chenhcs/FRoGS
6
- - Original model/code URL: https://github.com/chenhcs/FRoGS
7
 
8
  ## Public files
9
 
@@ -36,7 +34,7 @@
36
 
37
  This embedding represents genes using the official FRoGS ARCHS4 functional expression component.
38
 
39
- FRoGS vectors are functional gene representations. The ARCHS4 component is based on large-scale expression/co-expression information, so this embedding should be interpreted as an expression-derived functional representation rather than a sequence or literature embedding.
40
 
41
  ### Provenance fields
42
 
@@ -44,8 +42,6 @@ FRoGS vectors are functional gene representations. The ARCHS4 component is based
44
  original_model_or_method: FRoGS-ARCHS4-256
45
  input_data_type: official FRoGS precomputed gene_vec_archs4_256.csv table
46
  raw_source: ~/data/raw_embeddings/frogs_archs4_256_v1_1/gene_vec_archs4_256.csv
47
- processing_script: ~/scripts/process_embeddings/process_frogs_archs4_256_v1_1.py
48
- mapping_or_qc_report: ~/reports/mapping_reports/frogs_archs4_256_v1_1_mapping_report.txt
49
  source_download_page: https://github.com/chenhcs/FRoGS
50
  source_code_page: https://github.com/chenhcs/FRoGS
51
  source_identifier_type: Entrez Gene ID
@@ -60,13 +56,14 @@ FRoGS vectors are functional gene representations. The ARCHS4 component is based
60
 
61
  The FRoGS source vectors were not retrained locally. The retained source was the official precomputed FRoGS `gene_vec_archs4_256.csv` table.
62
 
63
- The main local decisions were about safe identifier mapping and duplicate handling. Source identifiers were Entrez Gene IDs. Because Entrez IDs can sometimes map imperfectly to the current Ensembl-centered master table, the final workflow used strict Entrez mapping and excluded ambiguous cases. A previous duplicate-Ensembl QC issue was fixed by collapsing only after mapping and averaging duplicate source rows that safely referred to the same final gene.
 
64
 
65
  ### Mapping and duplicate-handling decisions
66
 
67
  Source identifiers were parsed as Entrez Gene IDs and mapped to master_gene_table.csv through enriched Entrez fields. Entrez IDs mapping to multiple possible master-table genes were excluded as ambiguous. Multiple Entrez IDs or duplicate source rows mapping to the same final Ensembl gene were averaged after mapping.
68
 
69
- This preserves the rule used across the library: every final row must correspond to exactly one master-table gene, and every embedding must have one vector per retained gene.
70
 
71
  ### Processing workflow
72
 
@@ -84,7 +81,7 @@ For this embedding:
84
 
85
  This README describes the standardized gene-level version included in the public resource. 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.
86
 
87
- For FRoGS, the original biological representation was preserved from the official precomputed `gene_vec_archs4_256.csv` source table. The local work did not retrain FRoGS; it performed strict Entrez mapping, ambiguous-ID exclusion, duplicate averaging after mapping, QC, and conversion to the shared public format.
88
 
89
  ### Summary of final mapping:
90
 
 
1
  # Embedding name: FRoGS-ARCHS4-256
2
 
3
  - Modality: Co-expression / ARCHS4 functional expression embedding
 
4
  - Embedding source URL: https://github.com/chenhcs/FRoGS
 
5
 
6
  ## Public files
7
 
 
34
 
35
  This embedding represents genes using the official FRoGS ARCHS4 functional expression component.
36
 
37
+ FRoGS vectors are functional gene representations. The ARCHS4 component is based on large-scale expression/co-expression information, so this embedding should be interpreted as an expression-derived functional representation.
38
 
39
  ### Provenance fields
40
 
 
42
  original_model_or_method: FRoGS-ARCHS4-256
43
  input_data_type: official FRoGS precomputed gene_vec_archs4_256.csv table
44
  raw_source: ~/data/raw_embeddings/frogs_archs4_256_v1_1/gene_vec_archs4_256.csv
 
 
45
  source_download_page: https://github.com/chenhcs/FRoGS
46
  source_code_page: https://github.com/chenhcs/FRoGS
47
  source_identifier_type: Entrez Gene ID
 
56
 
57
  The FRoGS source vectors were not retrained locally. The retained source was the official precomputed FRoGS `gene_vec_archs4_256.csv` table.
58
 
59
+ The main local decisions were about safe identifier mapping and duplicate handling. Source identifiers were Entrez Gene IDs. Because Entrez IDs can sometimes map imperfectly to the current Ensembl-centered master table, the final workflow used strict Entrez mapping and excluded ambiguous cases.
60
+
61
 
62
  ### Mapping and duplicate-handling decisions
63
 
64
  Source identifiers were parsed as Entrez Gene IDs and mapped to master_gene_table.csv through enriched Entrez fields. Entrez IDs mapping to multiple possible master-table genes were excluded as ambiguous. Multiple Entrez IDs or duplicate source rows mapping to the same final Ensembl gene were averaged after mapping.
65
 
66
+ Every final row corresponds to exactly one master-table gene, and every embedding has one vector per retained gene.
67
 
68
  ### Processing workflow
69
 
 
81
 
82
  This README describes the standardized gene-level version included in the public resource. 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.
83
 
84
+ For FRoGS, the original biological representation was preserved from the official precomputed `gene_vec_archs4_256.csv` source table.
85
 
86
  ### Summary of final mapping:
87