Pasta

Index

1. Instantiate model class

2. Define clock metadata

3. Download clock dependencies

4. Load features

5. Load weights into base model

6. Load reference values

7. Load preprocess and postprocess objects

8. Check all clock parameters

9. Basic test

10. Save torch model

11. Clear directory

Let’s first import some packages:

import os
import inspect
import shutil
import json
import torch
import pandas as pd
import pyaging as pya

Instantiate model class

def print_entire_class(cls):
    source = inspect.getsource(cls)
    print(source)

print_entire_class(pya.models.Pasta)

class Pasta(pyagingModel):
    def __init__(self):
        super().__init__()

    @staticmethod
    def _rank_average(values):
        """
        Assign average ranks (1-based) per vector, handling ties.
        """
        sorted_vals, sorted_idx = torch.sort(values)
        ranks = torch.empty_like(sorted_vals, dtype=values.dtype)

        n = values.numel()
        start = 0
        while start < n:
            end = start + 1
            while end < n and sorted_vals[end] == sorted_vals[start]:
                end += 1
            avg_rank = (start + end - 1) / 2.0 + 1.0
            ranks[sorted_idx[start:end]] = avg_rank
            start = end

        return ranks

    def preprocess(self, x):
        """
        Fill missing values with the global median then rank-normalize per sample.
        """
        median = torch.nanmedian(x)
        if torch.isnan(median):
            median = torch.tensor(0.0, device=x.device, dtype=x.dtype)
        x = torch.where(torch.isnan(x), median, x)

        ranked = torch.empty_like(x, dtype=x.dtype)
        for i in range(x.size(0)):
            ranked[i] = self._rank_average(x[i])

        return ranked

    def postprocess(self, x):
        """
        Apply linear scaling and shifting constants from the original Pasta definition.
        """
        scale = self.postprocess_dependencies[0]
        offset_factor = self.postprocess_dependencies[1]
        return x * scale + offset_factor * scale

model = pya.models.Pasta()

Define clock metadata

model.metadata["clock_name"] = 'pasta'
model.metadata["data_type"] = 'transcriptomics'
model.metadata["species"] = 'Homo sapiens'
model.metadata["year"] = 2025
model.metadata["approved_by_author"] = '✅'
model.metadata["citation"] = "Salignon, Jerome, et al. \"Pasta, an age-shift transcriptomic clock, maps the chemical and genetic determinants of aging and rejuvenation.\" bioRxiv (2025): 2025-06."
model.metadata["doi"] = "https://doi.org/10.1101/2025.06.04.657785"
model.metadata["research_only"] = None
model.metadata["notes"] = "Rank-normalized transcriptomic clock."

Download clock dependencies

Download coefficient file

coeff_url = "https://raw.githubusercontent.com/bio-learn/biolearn/master/biolearn/data/Pasta.csv"
os.system(f"curl -L {coeff_url} -o Pasta.csv")

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  322k  100  322k    0     0  1174k      0 --:--:-- --:--:-- --:--:-- 1176k

0

Load features

From CSV file

coeffs = pd.read_csv('Pasta.csv')
coeffs['feature'] = coeffs['GeneID']
coeffs['coefficient'] = coeffs['CoefficientTraining']

model.features = coeffs['feature'].tolist()

Load weights into base model

From CSV file

weights = torch.tensor(coeffs['coefficient'].tolist()).unsqueeze(0)
intercept = torch.tensor([0.0])

Linear model

base_model = pya.models.LinearModel(input_dim=len(model.features))

base_model.linear.weight.data = weights.float()
base_model.linear.bias.data = intercept.float()

model.base_model = base_model

Load reference values

model.reference_values = [float("nan")] * len(model.features)

Load preprocess and postprocess objects

model.preprocess_name = "median_fill_and_rank_normalization"
model.preprocess_dependencies = None

model.postprocess_name = "scale_and_shift"
model.postprocess_dependencies = [-4.76348378687217, -0.0502893445253186]

Check all clock parameters

pya.utils.print_model_details(model)

%==================================== Model Details ====================================%
Model Attributes:

training: True
metadata: {'approved_by_author': '✅',
 'citation': 'Salignon, Jerome, et al. "Pasta, an age-shift transcriptomic '
             'clock, maps the chemical and genetic determinants of aging and '
             'rejuvenation." bioRxiv (2025): 2025-06.',
 'clock_name': 'pasta',
 'data_type': 'transcriptomics',
 'doi': 'https://doi.org/10.1101/2025.06.04.657785',
 'notes': 'Rank-normalized transcriptomic clock.',
 'research_only': None,
 'species': 'Homo sapiens',
 'version': None,
 'year': 2025}
reference_values: [nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan, nan]... [Total elements: 8113]
preprocess_name: 'median_fill_and_rank_normalization'
preprocess_dependencies: None
postprocess_name: 'scale_and_shift'
postprocess_dependencies: [-4.76348378687217, -0.0502893445253186]
features: ['ENSG00000196839', 'ENSG00000170558', 'ENSG00000133997', 'ENSG00000168060', 'ENSG00000101473', 'ENSG00000136754', 'ENSG00000113552', 'ENSG00000177485', 'ENSG00000136560', 'ENSG00000094631', 'ENSG00000108840', 'ENSG00000170248', 'ENSG00000153094', 'ENSG00000159921', 'ENSG00000165879', 'ENSG00000135451', 'ENSG00000142892', 'ENSG00000179776', 'ENSG00000167670', 'ENSG00000129484', 'ENSG00000041880', 'ENSG00000113361', 'ENSG00000141198', 'ENSG00000100284', 'ENSG00000013619', 'ENSG00000010017', 'ENSG00000105993', 'ENSG00000113810', 'ENSG00000182963', 'ENSG00000126261']... [Total elements: 8113]
base_model_features: None

%==================================== Model Details ====================================%
Model Structure:

base_model: LinearModel(
  (linear): Linear(in_features=8113, out_features=1, bias=True)
)

%==================================== Model Details ====================================%
Model Parameters and Weights:

base_model.linear.weight: [-2.4399256290053017e-05, -1.774273368937429e-05, 1.554851587570738e-05, 1.1031659596483223e-05, 1.6993128156173043e-05, 3.9308954001171514e-05, -0.00012627331307157874, 2.8949250463483622e-06, -6.271281017689034e-05, 2.9893646569689736e-05, 3.6174697015667334e-05, 6.864466558909044e-05, -2.3814825908630155e-05, 3.11008479911834e-05, 1.0880126865231432e-05, 9.605172635929193e-06, 1.1990639904979616e-05, 9.29949510464212e-06, 6.331568147288635e-05, -3.362866482348181e-05, -0.00022874546993989497, -2.7509766368893906e-05, 6.674586074950639e-06, 1.986255301744677e-05, -3.5506527638062835e-05, 2.922421663242858e-05, -4.5067787141306326e-05, 5.991863872623071e-05, 3.728850060724653e-05, 4.235586311551742e-05]... [Tensor of shape torch.Size([1, 8113])]
base_model.linear.bias: tensor([0.])

%==================================== Model Details ====================================%

Basic test

torch.manual_seed(42)
input = torch.randn(10, len(model.features), dtype=float)
model.eval()
model.to(float)
pred = model(input)
pred

tensor([[-12.0950],
        [-36.7366],
        [  2.5806],
        [ 29.9285],
        [-21.5258],
        [ -3.6489],
        [-39.3746],
        [ 36.6380],
        [ 27.4081],
        [-33.9218]], dtype=torch.float64, grad_fn=<AddBackward0>)

Save torch model

torch.save(model, f"../weights/{model.metadata['clock_name']}.pt")

Clear directory

# Function to remove a folder and all its contents
def remove_folder(path):
    try:
        shutil.rmtree(path)
        print(f"Deleted folder: {path}")
    except Exception as e:
        print(f"Error deleting folder {path}: {e}")

# Get a list of all files and folders in the current directory
all_items = os.listdir('.')

# Loop through the items
for item in all_items:
    # Check if it's a file and does not end with .ipynb
    if os.path.isfile(item) and not item.endswith('.ipynb'):
        os.remove(item)
        print(f"Deleted file: {item}")
    # Check if it's a folder
    elif os.path.isdir(item):
        remove_folder(item)

Deleted file: EPIC_salas_18_reference.csv
Deleted file: Pasta.csv