DNAmFitAge#

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:

[1]:

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

Instantiate model class#

[2]:

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

print_entire_class(pya.models.DNAmFitAge)

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

        self.GaitF = None
        self.GripF = None
        self.GaitM = None
        self.GripM = None
        self.VO2Max = None

        self.features_GaitF = None
        self.features_GripF = None
        self.features_GaitM = None
        self.features_GripM = None
        self.features_VO2Max = None

    def forward(self, x):
        Female = x[:, -3]  # .unsqueeze(1)
        Age = x[:, -2]  # .unsqueeze(1)
        GrimAge = x[:, -1].unsqueeze(1)

        is_female = Female == 1
        is_male = Female == 0

        x_f = x[is_female]
        x_m = x[is_male]

        GaitF = self.GaitF(x_f[:, self.features_GaitF])
        GripF = self.GripF(x_f[:, self.features_GripF])
        VO2MaxF = self.VO2Max(x_f[:, self.features_VO2Max])
        GrimAgeF = GrimAge[is_female, :]

        GaitM = self.GaitM(x_m[:, self.features_GaitM])
        GripM = self.GripM(x_m[:, self.features_GripM])
        VO2MaxM = self.VO2Max(x_m[:, self.features_VO2Max])
        GrimAgeM = GrimAge[is_male, :]

        x_f = torch.concat(
            [
                (VO2MaxF - 46.825091) / (-0.13620215),
                (GripF - 39.857718) / (-0.22074456),
                (GaitF - 2.508547) / (-0.01245682),
                (GrimAgeF - 7.978487) / (0.80928530),
            ],
            dim=1,
        )

        x_m = torch.concat(
            [
                (VO2MaxM - 49.836389) / (-0.141862925),
                (GripM - 57.514016) / (-0.253179827),
                (GaitM - 2.349080) / (-0.009380061),
                (GrimAgeM - 9.549733) / (0.835120557),
            ],
            dim=1,
        )

        y_f = self.base_model_f(x_f)
        y_m = self.base_model_m(x_m)

        y = torch.zeros((x.size(0), 1), dtype=x.dtype, device=x.device)
        y[is_female] = y_f
        y[is_male] = y_m

        return y

    def preprocess(self, x):
        return x

    def postprocess(self, x):
        return x


[3]:

model = pya.models.DNAmFitAge()

Define clock metadata#

[4]:

model.metadata["clock_name"] = "dnamfitage"
model.metadata["data_type"] = "DNA methylation"  # Paper: Blood DNA methylation was used to develop the fitness biomarkers.
model.metadata["species"] = "Homo sapiens"  # Paper: The development cohorts were human adult studies (FHS, BLSA, and Budapest).
model.metadata["year"] = 2023
model.metadata["approved_by_author"] = "⌛"
model.metadata["citation"] = "McGreevy, K. M., et al. “DNAmFitAge: biological age indicator incorporating physical fitness.” Aging 15(10): 3904–3938 (2023)."
model.metadata["doi"] = "https://doi.org/10.18632/aging.204538"
model.metadata["notes"] = "Sex-specific Klemera–Doubal biological-age composite combining DNAm gait speed, grip strength, VO2max, and DNAmGrimAge."
model.metadata["research_only"] = None
model.metadata["tissue"] = ["whole blood"]  # Paper: The biomarkers were developed from blood DNAm data.
model.metadata["predicts"] = ["physical-fitness biological age"]  # Paper: DNAmFitAge provides an estimate of biological age that incorporates fitness.
model.metadata["training_target"] = ["biological age"]  # Paper: The Klemera–Doubal framework posits an unobserved biological-age trait centered on chronological age.
model.metadata["unit"] = ["years"]  # Paper: FitAgeAcceleration is interpreted as years older or younger than expected chronological age.
model.metadata["model_type"] = "Klemera–Doubal composite"  # Paper: The TrueTrait function carried out the Klemera–Doubal method with sex-specific component weights.
model.metadata["platform"] = ["Illumina 450K"]  # Paper: The reported fitness CpG background and fitted loci were on the 450K array.
model.metadata["population"] = "adults"  # Paper: DNAmFitAge was formed in FHS and BLSA adults and combines sex-appropriate female and male fitness-component estimates.
model.metadata["journal"] = "Aging"
model.metadata["last_author"] = "Steve Horvath"
model.metadata["n_features"] = 630
model.metadata["citations"] = 99
model.metadata["citations_date"] = "2026-07-05"

Download clock dependencies#

Download GitHub repository#

[5]:

github_url = "https://github.com/kristenmcgreevy/DNAmFitAge.git"
github_folder_name = github_url.split('/')[-1].split('.')[0]
os.system(f"git clone {github_url}")

[5]:

0

Download from R package#

[6]:

%%writefile download.r

options(repos = c(CRAN = "https://cloud.r-project.org/"))
library(jsonlite)

DNAmFitnessModels <- readRDS("DNAmFitAge/DNAmFitnessModelsandFitAge_Oct2022.rds")

AllCpGs <- DNAmFitnessModels$AllCpGs
write_json(AllCpGs, "AllCpGs.json")

MaleMedians <- DNAmFitnessModels$Male_Medians_All
write.csv(MaleMedians, "MaleMedians.csv")
FemaleMedians <- DNAmFitnessModels$Female_Medians_All
write.csv(FemaleMedians, "FemaleMedians.csv")

Gait_noAge_Females <- DNAmFitnessModels$Gait_noAge_Females
Gait_noAge_Males <- DNAmFitnessModels$Gait_noAge_Males
Grip_noAge_Females <- DNAmFitnessModels$Grip_noAge_Females
Grip_noAge_Males <- DNAmFitnessModels$Grip_noAge_Males
VO2maxModel <- DNAmFitnessModels$VO2maxModel
write.csv(Gait_noAge_Females, "Gait_noAge_Females.csv")
write.csv(Gait_noAge_Males, "Gait_noAge_Males.csv")
write.csv(Grip_noAge_Females, "Grip_noAge_Females.csv")
write.csv(Grip_noAge_Males, "Grip_noAge_Males.csv")
write.csv(VO2maxModel, "VO2maxModel.csv")

Writing download.r

[7]:

os.system("Rscript download.r")

[7]:

0

Load features#

From JSON file#

[8]:

with open('AllCpGs.json', 'r') as f:
    features_list = json.load(f)
model.features = features_list + ['female'] + ['age'] + ['grimage']

Load weights into base model#

From CSV file#

[9]:

gaitf_df = pd.read_csv('Gait_noAge_Females.csv', index_col=0)
gaitm_df = pd.read_csv('Gait_noAge_Males.csv', index_col=0)
gripf_df = pd.read_csv('Grip_noAge_Females.csv', index_col=0)
gripm_df = pd.read_csv('Grip_noAge_Males.csv', index_col=0)
vo2max_df = pd.read_csv('VO2maxModel.csv', index_col=0)

Linear model#

[10]:

all_features = features_list + ['Female'] + ['Age'] + ['GrimAge']

model.GaitF = pya.models.LinearModel(input_dim=len(gaitf_df))
model.GaitF.linear.weight.data = torch.tensor(np.array(gaitf_df['estimate'][1:])).unsqueeze(0).float()
model.GaitF.linear.bias.data = torch.tensor(np.array(gaitf_df['estimate'].iloc[0])).float()
model.features_GaitF = torch.tensor([all_features.index(item) for item in np.array(gaitf_df['term'][1:]) if item in all_features]).long()

model.GaitM = pya.models.LinearModel(input_dim=len(gaitm_df))
model.GaitM.linear.weight.data = torch.tensor(np.array(gaitm_df['estimate'][1:])).unsqueeze(0).float()
model.GaitM.linear.bias.data = torch.tensor(np.array(gaitm_df['estimate'].iloc[0])).float()
model.features_GaitM = torch.tensor([all_features.index(item) for item in np.array(gaitm_df['term'][1:]) if item in all_features]).long()

model.GripF = pya.models.LinearModel(input_dim=len(gripf_df))
model.GripF.linear.weight.data = torch.tensor(np.array(gripf_df['estimate'][1:])).unsqueeze(0).float()
model.GripF.linear.bias.data = torch.tensor(np.array(gripf_df['estimate'].iloc[0])).float()
model.features_GripF = torch.tensor([all_features.index(item) for item in np.array(gripf_df['term'][1:]) if item in all_features]).long()

model.GaitM = pya.models.LinearModel(input_dim=len(gaitm_df))
model.GaitM.linear.weight.data = torch.tensor(np.array(gaitm_df['estimate'][1:])).unsqueeze(0).float()
model.GaitM.linear.bias.data = torch.tensor(np.array(gaitm_df['estimate'].iloc[0])).float()
model.features_GaitM = torch.tensor([all_features.index(item) for item in np.array(gaitm_df['term'][1:]) if item in all_features]).long()

model.GripM = pya.models.LinearModel(input_dim=len(gripm_df))
model.GripM.linear.weight.data = torch.tensor(np.array(gripm_df['estimate'][1:])).unsqueeze(0).float()
model.GripM.linear.bias.data = torch.tensor(np.array(gripm_df['estimate'].iloc[0])).float()
model.features_GripM = torch.tensor([all_features.index(item) for item in np.array(gripm_df['term'][1:]) if item in all_features]).long()

model.VO2Max = pya.models.LinearModel(input_dim=len(vo2max_df))
model.VO2Max.linear.weight.data = torch.tensor(np.array(vo2max_df['estimate'][1:])).unsqueeze(0).float()
model.VO2Max.linear.bias.data = torch.tensor(np.array(vo2max_df['estimate'].iloc[0])).float()
model.features_VO2Max = torch.tensor([all_features.index(item) for item in np.array(vo2max_df['term'][1:]) if item in all_features]).long()

Linear model#

[11]:

base_model_m = pya.models.LinearModel(input_dim=4)

base_model_m.linear.weight.data = torch.tensor(np.array([0.1390346, 0.1787371, 0.1593873, 0.5228411])).unsqueeze(0).float()
base_model_m.linear.bias.data = torch.tensor(np.array([0.0])).float()

model.base_model_m = base_model_m

base_model_f = pya.models.LinearModel(input_dim=4)

base_model_f.linear.weight.data = torch.tensor(np.array([0.1044232, 0.1742083, 0.2278776, 0.4934908])).unsqueeze(0).float()
base_model_f.linear.bias.data = torch.tensor(np.array([0.0])).float()

model.base_model_f = base_model_f

Load reference values#

From CSV file#

[12]:

reference_df_f = pd.read_csv('FemaleMedians.csv', index_col=0)
reference_f = reference_df_f.loc[1, model.features[:-3]]
reference_df_m = pd.read_csv('MaleMedians.csv', index_col=0)
reference_m = reference_df_m.loc[1, model.features[:-3]]
reference = (reference_f + reference_m)/2
model.reference_values =  list(reference) + [1] + [65] + [65] #65yo F with 65GrimAge

Load preprocess and postprocess objects#

[13]:

model.preprocess_name = None
model.preprocess_dependencies = None

[14]:

model.postprocess_name = None
model.postprocess_dependencies = None

Check all clock parameters#

[15]:

pya.utils.print_model_details(model)


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

training: True
metadata: {'approved_by_author': '⌛',
 'citation': 'McGreevy, Kristen M., et al. "DNAmFitAge: biological age '
             'indicator incorporating physical fitness." Aging (Albany NY) '
             '15.10 (2023): 3904.',
 'clock_name': 'dnamfitage',
 'data_type': 'methylation',
 'doi': 'https://doi.org/10.18632/aging.204538',
 'notes': 'Reference values is mean between male and female training medians',
 'research_only': None,
 'species': 'Homo sapiens',
 'version': None,
 'year': 2023}
reference_values: [0.521913219528255, 0.28125954210819004, 0.927523008548927, 0.01455467410155745, 0.041014116925727054, 0.12647568639954998, 0.7148617994059816, 0.6786637301838809, 0.909376031310397, 0.1136806555747305, 0.45398237911395245, 0.0544492346267719, 0.7738429377348031, 0.8480746411296824, 0.7667083937960659, 0.0159858833215953, 0.7183128068669931, 0.06813828137044395, 0.939547714031041, 0.8290646522154059, 0.01727972597475225, 0.0697125677059708, 0.366626793673691, 0.588925514102081, 0.027865661596066897, 0.8252930680510391, 0.211681997462417, 0.01269953071843695, 0.7886011964686286, 0.871255311509148]... [Total elements: 630]
preprocess_name: None
preprocess_dependencies: None
postprocess_name: None
postprocess_dependencies: None
features: ['cg25137787', 'cg08911391', 'cg24685778', 'cg25551287', 'cg15322207', 'cg24604749', 'cg03890680', 'cg14601038', 'cg25587481', 'cg23922134', 'cg18561976', 'cg08175029', 'cg23202468', 'cg06181470', 'cg14422932', 'cg15128470', 'cg13587180', 'cg25440680', 'cg16995193', 'cg12864235', 'cg23715435', 'cg02805890', 'cg06381959', 'cg00779476', 'cg25489467', 'cg02376916', 'cg17741339', 'cg20219159', 'cg19629631', 'cg25577212']... [Total elements: 630]
base_model_features: None
base_model: None
features_GaitF: [272, 504, 505, 506, 507, 508, 509, 510, 511, 512, 513, 281, 514, 515, 516, 517, 518, 301, 519, 305, 520, 521, 522, 523, 524, 525, 526, 314, 318, 527]... [Tensor of shape torch.Size([53])]
features_GripF: [272, 273, 274, 275, 276, 277, 278, 279, 280, 281, 282, 283, 284, 285, 286, 287, 288, 289, 290, 291, 292, 293, 294, 295, 296, 297, 298, 299, 300, 301]... [Tensor of shape torch.Size([91])]
features_GaitM: [545, 546, 481, 547, 548, 483, 549, 550, 485, 551, 486, 148, 552, 553, 554, 488, 149, 555, 556, 557, 558, 559, 234, 560, 561, 562, 563, 564, 465, 565]... [Tensor of shape torch.Size([59])]
features_GripM: [361, 362, 363, 209, 364, 365, 366, 367, 211, 368, 369, 370, 371, 372, 373, 374, 375, 376, 377, 378, 379, 380, 381, 382, 383, 384, 385, 386, 387, 225]... [Tensor of shape torch.Size([93])]
features_VO2Max: [587, 588, 589, 590, 591, 592, 593, 594, 595, 596, 597, 598, 599, 600, 601, 602, 603, 604, 605, 606, 607, 608, 609, 610, 611, 612, 613, 614, 615, 616]... [Tensor of shape torch.Size([41])]

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

GaitF: LinearModel(
  (linear): Linear(in_features=54, out_features=1, bias=True)
)
GaitM: LinearModel(
  (linear): Linear(in_features=60, out_features=1, bias=True)
)
GripF: LinearModel(
  (linear): Linear(in_features=92, out_features=1, bias=True)
)
GripM: LinearModel(
  (linear): Linear(in_features=94, out_features=1, bias=True)
)
VO2Max: LinearModel(
  (linear): Linear(in_features=42, out_features=1, bias=True)
)
base_model_m: LinearModel(
  (linear): Linear(in_features=4, out_features=1, bias=True)
)
base_model_f: LinearModel(
  (linear): Linear(in_features=4, out_features=1, bias=True)
)

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

GaitF.linear.weight: [-0.05867812782526016, -0.05637867748737335, -0.10371068120002747, 0.01624305173754692, -0.053210534155368805, -0.07633326947689056, -0.01514248363673687, -0.049918416887521744, -0.013779371976852417, 0.14702405035495758, 0.22061608731746674, -0.6326411366462708, -0.40431228280067444, 0.06633924692869186, -0.2228449434041977, -0.03177845478057861, -0.35903501510620117, 0.4153103232383728, -0.837234616279602, 0.056484829634428024, -0.13299566507339478, -0.058516617864370346, 0.04777200520038605, 0.13982263207435608, -0.1280703842639923, -0.03444225341081619, -0.05433110147714615, -0.4258767366409302, 0.0011224570916965604, 0.01846371404826641]... [Tensor of shape torch.Size([1, 53])]
GaitF.linear.bias: tensor(3.9701)
GaitM.linear.weight: [0.10235995799303055, 0.08753516525030136, 0.3124901354312897, -0.28120002150535583, -0.3208324611186981, 0.24479524791240692, 0.05682919919490814, 0.21363066136837006, -0.3853186368942261, -0.038501303642988205, -0.0023554968647658825, -0.17415688931941986, 0.05159717798233032, -0.5185700058937073, -0.04655730724334717, -0.19074112176895142, -0.21096128225326538, 0.011959427036345005, 0.1078566312789917, 0.0770212784409523, 0.18820391595363617, 0.43347951769828796, -0.13240143656730652, 0.021351546049118042, -0.12319610267877579, -0.010150707326829433, -0.007736711762845516, 0.13240836560726166, -1.1829639673233032, -0.10984379798173904]... [Tensor of shape torch.Size([1, 59])]
GaitM.linear.bias: tensor(3.1825)
GripF.linear.weight: [-2.3457672595977783, -2.94146728515625, 2.8132119178771973, -0.9427763223648071, 3.2280826568603516, -0.5796002745628357, -5.075088977813721, -1.3516250848770142, 3.539742946624756, -6.468724727630615, -3.5424692630767822, -6.332897663116455, 4.4002580642700195, 10.170988082885742, -0.5222252011299133, -2.993544101715088, -0.7089398503303528, -3.3968186378479004, 0.9145923852920532, 1.0081183910369873, -2.5558736324310303, -1.6970638036727905, 2.0081098079681396, 0.2233070731163025, -3.5272421836853027, -4.740792274475098, -2.4629898071289062, 0.7111413478851318, -11.599475860595703, 3.976231575012207]... [Tensor of shape torch.Size([1, 91])]
GripF.linear.bias: tensor(53.8206)
GripM.linear.weight: [-0.892844021320343, 1.6379542350769043, 1.5265462398529053, -0.3347032964229584, -1.9029316902160645, -0.2647155225276947, -6.30814266204834, 14.954833984375, 1.178484320640564, 3.5211784839630127, -0.1861504316329956, -1.6255935430526733, 4.550158977508545, -1.587499976158142, -0.449662446975708, -8.599822998046875, 25.895660400390625, 4.368823051452637, 3.992393970489502, 1.3252184391021729, -2.2360410690307617, -0.6896253228187561, 1.5932470560073853, 1.5443568229675293, -0.7052236795425415, -3.0787854194641113, -0.2242996096611023, -0.23673297464847565, 2.1442930698394775, -0.3954241871833801]... [Tensor of shape torch.Size([1, 93])]
GripM.linear.bias: tensor(43.0198)
VO2Max.linear.weight: [5.249129772186279, 3.0901758670806885, -7.551166534423828, -5.796545028686523, -1.094834804534912, -2.3806116580963135, -0.0022889631800353527, 1.0938740968704224, -1.4775551557540894, 1.4427802562713623, 1.268430471420288, 5.4764933586120605, -8.934550285339355, -1.9918478727340698, -5.6620774269104, -6.2174201011657715, -0.6082701086997986, -7.513339996337891, -1.4299590587615967, -3.6723220348358154, 14.669830322265625, 0.5884844660758972, -0.9597266912460327, -1.0253041982650757, -1.802089810371399, -4.9922356605529785, -0.6746888160705566, -10.973499298095703, -0.6614307761192322, 2.365175247192383]... [Tensor of shape torch.Size([1, 41])]
VO2Max.linear.bias: tensor(69.6523)
base_model_m.linear.weight: tensor([[0.1390, 0.1787, 0.1594, 0.5228]])
base_model_m.linear.bias: tensor([0.])
base_model_f.linear.weight: tensor([[0.1044, 0.1742, 0.2279, 0.4935]])
base_model_f.linear.bias: tensor([0.])

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

Basic test#

[16]:

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

[16]:

tensor([[0.],
        [0.],
        [0.],
        [0.],
        [0.],
        [0.],
        [0.],
        [0.],
        [0.],
        [0.]], dtype=torch.float64, grad_fn=<IndexPutBackward0>)

Save torch model#

[17]:

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

Clear directory#

[18]:

# 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: Grip_noAge_Females.csv
Deleted file: Grip_noAge_Males.csv
Deleted file: Gait_noAge_Females.csv
Deleted file: VO2maxModel.csv
Deleted file: AllCpGs.json
Deleted file: Gait_noAge_Males.csv
Deleted folder: DNAmFitAge
Deleted file: download.r
Deleted file: FemaleMedians.csv
Deleted file: MaleMedians.csv