causal snps | causal variants | tensorflow | 神经网络实战 | Data Simulation

先读几篇文章：

Interpretation of Association Signals and Identification of Causal Variants from Genome-wide Association Studies

GWAS have been successful in identifying disease susceptibility loci, but it remains a challenge to pinpoint the causal variants in subsequent fine-mapping studies. A conventional fine-mapping effort starts by sequencing dozens of randomly selected samples at susceptibility loci to discover candidate variants, which are then placed on custom arrays or used in imputation algorithms to find the causal variants. We propose that one or several rare or low-frequency causal variants can hitchhike the same common tag SNP, so causal variants may not be easily unveiled by conventional efforts. Here, we first demonstrate that the true effect size and proportion of variance explained by a collection of rare causal variants can be underestimated by a common tag SNP, thereby accounting for some of the “missing heritability” in GWAS. We then describe a case-selection approach based on phasing long-range haplotypes and sequencing cases predicted to harbor causal variants. We compare this approach with conventional strategies on a simulated data set, and we demonstrate its advantages when multiple causal variants are present. We also evaluate this approach in a GWAS on hearing loss, where the most common causal variant has a minor allele frequency (MAF) of 1.3% in the general population and 8.2% in 329 cases. With our case-selection approach, it is present in 88% of the 32 selected cases (MAF = 66%), so sequencing a subset of these cases can readily reveal the causal allele. Our results suggest that thinking beyond common variants is essential in interpreting GWAS signals and identifying causal variants.

Where is the causal variant? On the advantage of the family design over the case-control design in genetic association studies.

Identification of causal genes for complex traits

Pure and Confounded Effects of Causal SNPs on Longevity: Insights for Proper Interpretation of Research Findings in GWAS of Populations with Different Genetic Structures

初步学习一些TensorFlow的基本概念

YouTube的莫凡教程  GitHub

# View more python tutorial on my Youtube and Youku channel!!!

# Youtube video tutorial: https://www.youtube.com/channel/UCdyjiB5H8Pu7aDTNVXTTpcg
# Youku video tutorial: http://i.youku.com/pythontutorial

"""
Please note, this code is only for python 3+. If you are using python 2+, please modify the code accordingly.
"""
from __future__ import print_function
import tensorflow as tf
import numpy as np

# create data
x_data = np.random.rand(100).astype(np.float32)
y_data = x_data*0.1 + 0.3

### create tensorflow structure start ###
Weights = tf.Variable(tf.random_uniform([1], -1.0, 1.0))
biases = tf.Variable(tf.zeros([1]))

y = Weights*x_data + biases

loss = tf.reduce_mean(tf.square(y-y_data))
optimizer = tf.train.GradientDescentOptimizer(0.5)
train = optimizer.minimize(loss)
### create tensorflow structure end ###

sess = tf.Session()
# tf.initialize_all_variables() no long valid from
# 2017-03-02 if using tensorflow >= 0.12
if int((tf.__version__).split('.')[1]) < 12 and int((tf.__version__).split('.')[0]) < 1:
    init = tf.initialize_all_variables()
else:
    init = tf.global_variables_initializer()
sess.run(init)

for step in range(201):
    sess.run(train)
    if step % 20 == 0:
        print(step, sess.run(Weights), sess.run(biases))

　　

如何制作模拟的数据

Data Simulation Software for Whole-Genome Association and Other Studies in Human Genetics

A comparison of tools for the simulation of genomic next-generation sequencing data

num_cau_SNP <- 20
num_SNP <- 500
samplesize <- 20
h_squared <- 0.5

# generate genotype in Binomial distribution
pj <- runif(num_SNP, 0.01, 0.5)
xij_star <- matrix(0, samplesize, num_SNP)
#for every SNP
for (j in 1: num_SNP)
{
  xij_star[,j] <- rbinom(samplesize, 2, pj[j])
}

#position of causal SNPs
CauSNP <- sample(1:num_SNP, num_cau_SNP, replace = F)
Ord_CauSNP <- sort(CauSNP, decreasing = F)

# generate beta, which is the best predictor
beta <- rep(0,num_SNP)
dim(beta) <- c(num_SNP,1)
# non-null betas follow standard normal distribution
beta[Ord_CauSNP] <- rnorm(num_cau_SNP,0,1)

# epsilon
var_e <- sum((xij_star %*% beta)^2)
# var_e <- t(beta)%*%t(xij_star)%*%xij_star%*%beta/samplesize*(1-h_squared)/h_squared
e <- rnorm(samplesize, 0,sqrt(var_e))
dim(e) <- c(samplesize, 1)

# generate phenotype
pheno <- xij_star %*% beta + e

# scale(genotype matrix)

　　

待续~