genome wide pertub-seq

 CRISPi + scRNA

https://gwps.wi.mit.edu/

matrix file in H5AD format

https://doi.org/10.25452/figshare.plus.20029387

Genome wide screen targeted n=9867 genes. 

Essential-wide screen targeted n=2285 essential genes. 

Growth phenotypes were measured log2-guide enrichment per cell doubling (gamma)

"The relative homogeneity of CRISPRi reduces selection for unperturbed cells, especially when studying essential genes.  Unlike CRISPR knockout, CRISPRi does not lead to activation of the DNA damage response which can alter transcriptional signatures (Haapaniemi et al., 2018)."

"We use a compact, multiplexed CRISPR interference (CRISPRi) library to assay thousands of loss-of-function genetic perturbations with single-cell RNA sequencing (scRNA-seq) in chronic myeloid leukemia (CML) (K562) and retinal pigment epithelial (RPE1) cell lines."

There are four datasets on SRA:

1. K562 day 8 Perturb-seq (KD8): targeting all expressed genes at day 8 after transduction
2. RPE1 day 7 Perturb-seq (RD7): targeting DepMap essential genes at day 7 after transduction
3. K562 day 6 Perturb-seq (KD7): targeting DepMap essential genes at day 6 after transduction
4. K562 day 8 Perturb-seq (KD8_ultima): scRNA-seq libraries from the KD8 experiment sequenced on the Ultima sequencing platform rather than the Illumina sequencing platform

Potential problem: essential gene cannot be deleted? 

This is good data set for graph controllability, graph neural network analysis. 

(BUILD) human gene network reliability and pathogenic association of genetic variations in human populations

Reliability of human gene networks and their pathogenic implications.

Reliability of gene network in different human tissues and cell types -> robustness, cancer incidence?

health disparity
aging associated genes

expression profiling, ngs to infer tissue specific gene networks.

human twin aging expression
http://genomebiology.com/2013/14/7/R75?utm_campaign=10_12_13_genomebiol_Article_Mailing_Reg&utm_content=7387379393&utm_medium=BMCemail&utm_source=Emailvision

gwa aging 

http://www.ncbi.nlm.nih.gov/pubmed/21782286

age of puberty in japanese

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0063821

CR effect on cell lines SNPS
flow cytometers
fluorecesnce microscope

yeast model?

stochasticity X average degee > 1 for cells to be alive.

# Notice that degree*p > 1 if the cells are to be alive. >2 should sufficient redundancy.

Sagi, Wolf, Koonin 2012, PlosComBiol, Universal pacemakers of genome evolution

SWK12 focus on 6901 orthologous gene families in 41 archaeal and 59 bacterial genomes, provided in supporting text.

So, if the SWK12 molecular pacemakers occur in yeast, they are very conserved genes.  Maybe, I could simulate the aging of the evolutionary core network, and then study how the recent network component change the aging dynamics of the core network.  This is a way to see how evolution influence network reliability, i.e., network robustness.

Derivative of R with respect to p, when $c$ is a constant

\begin{align}
\frac{dR}{dp} = cmn\lambda(1-p)^{n-1} + cmnp\lambda (n-1) (1-p)^{n-2}
\end{align}

When $c$ is treated as a constant, the mode is p = 1/n. Hence, when np>1, R decreases as p increases. 

The above inference does not consider that the normalization parameter $c$ also depends on $p$.

2013 Oct 3: $c$ is actually very close to 1 in most simulations.  So, $c$ can be considered as a constant in most cases. Besides, mode p=1/n clearly makes intuitive sense. 

Mean field approximation and network reliability

According to wikipedia entry, MFT simply the behavior of large and complex stochastic models by studying a simpler model. Such models often consists a large number of small interacting individuals.  The effect of all the other individual on any given individuals is approximated by "a single average effect", thus reducing the many-body problem to a one-body problem.

I should apply mean field approximation in network reliability studies. The challenge is that biological networks are heterogeneous, and simple 'average' might leave some interesting properties. In any case, this is an interesting direction that I should explore.

In Bialek, Nemenman, Tishby 2008, Predictability, complexity and learning, BNT08 discussed a Ising model as

BNT08 used Boltzmann distribution to describe spins {$\sigma_i$}.

The Boltzmann distribution is basically exponential decay function that is often used in reliability models. So, there seems to be a natural connection between statistics physics and reliability modeling.

There may be a problem or challenge for the mean field approach to study aging. Based on reliability model, system ages are determined by extreme values of components. So, mean field approximation may not capture this maximal-minimal nature of aging.










Reference:
http://en.wikipedia.org/wiki/Mean_field_theory