Correct publication references, input guidance and output units

README.md

@@ -9,7 +9,7 @@ tags:
 
 # cpgptgrimage3
 
-CpGPTGrimAge3 implementation combining chronological age, GrimAge2 DNAm proxies, and CpGPT-predicted plasma-protein proxies in a Cox linear predictor that is calibrated to years.
+CpGPTGrimAge3 implementation combining chronological age, GrimAge2 DNAm proxies, and CpGPT-predicted plasma-protein proxies in a Cox linear predictor that is calibrated to years. Supply derived methylation protein/lifestyle proxies and age, not a raw CpG matrix or measured proteomics. CpGPT protein proxies come from the proteins checkpoint on its trained standardized scale; required proxies must be present.
 
 Model weights retain the original authors' terms; the pyaging software license does not relicense them. These weights are restricted to research use under the authors' terms.
 

config.json

@@ -10,7 +10,7 @@
  "last_author": "Bo Wang",
  "model_type": "Cox proportional hazards regression",
  "n_features": 24,
- "notes": "CpGPTGrimAge3 implementation combining chronological age, GrimAge2 DNAm proxies, and CpGPT-predicted plasma-protein proxies in a Cox linear predictor that is calibrated to years.",
+ "notes": "CpGPTGrimAge3 implementation combining chronological age, GrimAge2 DNAm proxies, and CpGPT-predicted plasma-protein proxies in a Cox linear predictor that is calibrated to years. Supply derived methylation protein/lifestyle proxies and age, not a raw CpG matrix or measured proteomics. CpGPT protein proxies come from the proteins checkpoint on its trained standardized scale; required proxies must be present.",
  "platform": [
   "Illumina 450K"
  ],

cpgptgrimage3.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:15ddd9eb55ff9e43c9a8c6884f5a1c502d5189c47833ab4f549ce6b32d300af0
-size 7253
+oid sha256:7c5176232bc17fe53cc030bfc2e2fd4914f92fc2190199100fef63ef0d928981
+size 7549