GreenGenomicsLab commited on
Commit
fdebf97
·
verified ·
1 Parent(s): 3eca795

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +53 -20
README.md CHANGED
@@ -8,21 +8,24 @@ tags:
8
  - nanoGPT
9
  - metagenomics
10
  - tara-oceans
 
11
  pipeline_tag: text-classification
12
  ---
13
 
14
  # algaGPT
15
 
16
- Causal language model for binary classification of microalgal vs. contaminant protein sequences.
17
 
18
  ## Model Description
19
 
20
- - **Architecture:** nanoGPT (Andrej Karpathy)
21
- - **Task:** Binary classification of microalgal protein sequences via next-token prediction
22
- - **Mode:** TI-inclusive (full-length sequences)
23
- - **Training data:** ~58.6M protein sequences (1:1 algal:contaminant ratio)
24
- - **Algal sources:** 166 microalgal genomes across 10 phyla
25
- - **Contaminant sources:** Bacterial, archaeal, and fungal sequences from NCBI nr
 
 
26
 
27
  ## Performance
28
 
@@ -33,12 +36,30 @@ Causal language model for binary classification of microalgal vs. contaminant pr
33
 
34
  ## Usage
35
 
36
- **Input:** Protein sequence (amino acid string)
 
37
  **Output:** Classification tag (algal/contaminant) via next-token prediction
38
 
39
- ## Applications
 
 
 
 
 
40
 
41
- algaGPT was used as the primary proteome extraction tool in the ELF-NET study (Nelson et al., forthcoming), where it purified algal protein sequences from 2,044 TARA Oceans metagenome assemblies, yielding 221.9 million sequences for downstream domain-environment coupling analysis.
 
 
 
 
 
 
 
 
 
 
 
 
42
 
43
  ## Authors
44
 
@@ -48,19 +69,31 @@ Green Genomics Lab, New York University Abu Dhabi
48
 
49
  ## Citation
50
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
51
  ```bibtex
52
- @article{la4sr2025,
53
- title={Pan-microalgal dark proteome mapping via interpretable deep learning and synthetic chimeras},
54
- author={Nelson, David R. and Jaiswal, Ashish Kumar and Ismail, Noha Samir and Mystikou, Alexandra and Salehi-Ashtiani, Kourosh},
55
- journal={Patterns},
56
- volume={6},
57
- pages={101373},
58
- year={2025},
59
- publisher={Cell Press},
60
- doi={10.1016/j.patter.2025.101373}
61
  }
62
  ```
63
 
64
  ## Contact
65
 
66
- Kourosh Salehi-Ashtiani — ksa3@nyu.edu
 
8
  - nanoGPT
9
  - metagenomics
10
  - tara-oceans
11
+ - marine-biology
12
  pipeline_tag: text-classification
13
  ---
14
 
15
  # algaGPT
16
 
17
+ Causal language model for binary classification of microalgal vs. contaminant protein sequences, built on [nanoGPT](https://github.com/karpathy/nanoGPT). algaGPT distinguishes algal proteins from bacterial, archaeal, and fungal contaminants in metagenomic assemblies without requiring sequence homology.
18
 
19
  ## Model Description
20
 
21
+ | Property | Value |
22
+ |----------|-------|
23
+ | Architecture | nanoGPT (GPT-2, 12 layers, 12 heads, 768 embedding) |
24
+ | Task | Binary classification via next-token prediction |
25
+ | Mode | TI-inclusive (full-length amino acid sequences) |
26
+ | Training data | ~58.6M protein sequences (1:1 algal:contaminant) |
27
+ | Algal sources | 166 microalgal genomes across 10 phyla |
28
+ | Contaminant sources | Bacterial, archaeal, and fungal sequences from NCBI nr |
29
 
30
  ## Performance
31
 
 
36
 
37
  ## Usage
38
 
39
+ **Input:** Amino acid string ending with `>`
40
+
41
  **Output:** Classification tag (algal/contaminant) via next-token prediction
42
 
43
+ ```python
44
+ # See Nelson et al. (2025) for full inference pipeline
45
+ # Batch inference: https://github.com/SarahD4/dark-whiteGPLM/blob/main/batch_inference.py
46
+ ```
47
+
48
+ ## Application in ELF-NET
49
 
50
+ algaGPT served as the primary proteome extraction tool in the ELF-NET study, purifying algal protein sequences from 2,044 TARA Oceans metagenome assemblies to yield 221.9 million sequences for downstream domain-environment coupling analysis.
51
+
52
+ ## Related Resources
53
+
54
+ | Resource | Link |
55
+ |----------|------|
56
+ | ELF-NET analysis pipeline (371 scripts, 15 modules) | [github.com/olympus-terminal/ELF-NET](https://github.com/olympus-terminal/ELF-NET) |
57
+ | Bidirectional XGBoost models | [TARA-XGBoost-Bidirectional](https://huggingface.co/GreenGenomicsLab/TARA-XGBoost-Bidirectional) |
58
+ | VICReg joint embedding model | [TARA-WorldModel-VICReg](https://huggingface.co/GreenGenomicsLab/TARA-WorldModel-VICReg) |
59
+ | Dark-whiteGPLM checkpoints | [SarahDaakour/dark-whiteGPLM](https://huggingface.co/SarahDaakour/dark-whiteGPLM) |
60
+ | Dark-whiteGPLM training data | [SarahDaakour/dark-whiteGPLM-data](https://huggingface.co/datasets/SarahDaakour/dark-whiteGPLM-data) |
61
+ | Dark-whiteGPLM code | [github.com/SarahD4/dark-whiteGPLM](https://github.com/SarahD4/dark-whiteGPLM) |
62
+ | algaGPT-purified sequences (Data S2) | [Zenodo 10.5281/zenodo.18728837](https://doi.org/10.5281/zenodo.18728837) |
63
 
64
  ## Authors
65
 
 
69
 
70
  ## Citation
71
 
72
+ algaGPT was introduced in:
73
+
74
+ ```bibtex
75
+ @article{nelson2025la4sr,
76
+ title = {Pan-microalgal dark proteome mapping via interpretable deep learning and synthetic chimeras},
77
+ author = {Nelson, David R. and Jaiswal, Ashish Kumar and Ismail, Noha Samir and Mystikou, Alexandra and Salehi-Ashtiani, Kourosh},
78
+ journal = {Patterns},
79
+ volume = {6},
80
+ pages = {101373},
81
+ year = {2025},
82
+ doi = {10.1016/j.patter.2025.101373}
83
+ }
84
+ ```
85
+
86
+ algaGPT was applied at scale in:
87
+
88
  ```bibtex
89
+ @article{nelson2026elfnet,
90
+ title = {Coupling of oceanographic state to the dark proteome: a foundation for genome-informed marine productivity modeling},
91
+ author = {Nelson, David Roy and Plouviez, Maxence and Daakour, Sarah and Jaiswal, Ashish and Fu, Weiqi and Amin, Shady A. and Salehi-Ashtiani, Kourosh},
92
+ journal = {Forthcoming},
93
+ year = {2026}
 
 
 
 
94
  }
95
  ```
96
 
97
  ## Contact
98
 
99
+ Kourosh Salehi-Ashtiani -- ksa3@nyu.edu