Alzheimer's is not One Disease!

What % of academic papers make serious headway in resolving a question? In my opinion, not many. 1% is optimistic. Here's a paper that falls into that 1%: Molecular subtyping of Alzheimer’s disease using RNA sequencing data reveals novel mechanisms and targets. There's plenty of tricky bioinformatics within, but the basic task isn't hard to grok: get a load of Alzheimer's data from a particular region of the brain, cluster it, and use every tool you can find to link the clusters to patient types, cell types, mouse studies, whatever. 

The big takeaway is that Alzheimer's isn't a single entity. In fact, that's an understatement. The same genes that are upregulated relative to controls in Cluster A, for example, may be downregulated in Cluster C. These are not subtle differences between clusters...they're massive. 

There were 5 clusters in the study (A, B1, B2, C1, C2), each with characteristic up/downregulated transcripts. We took those 5 pairs of transcripts, passed them through our "Match Studies" app and, in a format that would never pass peer review, summarize the results below. We also passed the WhatIsMyGene canonical Alzheimer's regulation lists (up/down) through the app.

A
tau neighborhood genes, tau protein binding, up-regulation of gaba/glutaminergic, dendritic synaptic pathways, downregulation of immune response, increase of neuronal regulation genes,tauP301L model
mimics	counters
arx ko	C2
B2	C1
high activity	alcoholism
B1	autism
brains lacking b/NK cells	some alzheimer studies
	hras g12s mut
	rosmarinic acid
	hoxa5 gain
	IAV infection
	methylphenidate
	tyr-trp dipeptide
	GBM stem cells w/gpr56 ko
	arx mutant
	huntington's mouse
	med23 ko
B1
tau neighborhood genes, up-regulation of gaba/glutaminergic, glycinergic, dendritic synaptic pathways, downregulation of immune response, downregulation of oligodendrocytic genes, APOJ/CD2AP/BIN1 mouse models
mimics	counters
B2	C1
A	some alzheimer studies
B2
tau neighborhood genes, up-regulation of gaba/glutaminergic, glycinergic, dendritic synaptic pathways, immune pathways, APOE2 dosage, downregulation of oligodendrocytic genes, APOJ/CD2AP/BIN1 mouse models
mimics	counters
B1	some alzheimer studies
C1	idh1 expression in NSCs
C2
A
C1
AB binding, fiber clearance, down-regulation of gaba/glutaminergic, glycinergic, dendritic synaptic pathways, immune pathways, APOE4 dosage, 5XFAD/APP Dutch/APP Swedish mice
mimics	counters
C2	A
autism	B1
nasu hakola brain	antiretrovirals
alcoholics	high activity
HIV patients (w/o antiretrovirals)	schizophrenia
MS lesions	zinc restriction
cocaine addiction	DHA treatment
B2	apoe4 organoids
ALS motor cortex	high protein diet
Creutzfeldt-Jakob
Down's syndrome
JEV/WNV infection
wig1 kd
mog (neuroinflammation)
hoxa5 gain
HIV
depression
C2
down-regulation of gaba/glutaminergic, glycinergic, dendritic synaptic pathways, immune pathways, 5XFAD/APP Dutch/APP Swedish mice
mimics	counters
C1	A
alcoholism	arx ko
autism	sufu ko
nasu-hakola	antiretrovirals
cocaine addiction
B2
hiv
hoxa5 gain
ALS motor cortex
Creutzfeldt-Jakob
APP ko
Dicer ko
WhatIsMyGene canonical Alzheimer's
canonical mimics	canonical counters
Creutzfeldt-Jakob	antiretrovirals
alcoholism	rdm11 kd
cocaine addiction	high activity
nasu hakola brain	dha treatment
HIV	A
C2	ndp ko
C1	schizophrenia
ALS	some Alzheimer's studies
MS lesions	9thc treatment
autism	cga ko
Down's syndrome	bdnf treatment
hoxa5 gain	sdhd kd
aged brain	pten heterozygosity
Cockayne syndrome	arx ko
Rhett's syndrome	fluoxetine
glioma stem cells (vs normal)
bipolar
OCD
hras g12s mutation
mouse epilepsy
rosmarinic acid
fibrinogen treatment
arx mutation
traumatic injury
anti-cd8 treatment

First, you see the cluster names. Then, in gray boxes, you see some broad characteristics of the clusters that were mentioned in the paper. Below that, you see studies that either mimic the clusters or counteract them, listed in order of significance. Studies highlighted in yellow are Alzheimer's mimics with respect to one cluster, but counters with respect to another. Green highlights suggest interventions that aren't contradicted in one cluster versus another. If you want the finer grain details of these studies, you can use the "Match Studies" tool yourself: enter our database IDs for the up/downregulation results in the clusters ("signatures" is actually the term in the paper), choose "brain" as "Cell Type", and choose "Inverse Correlations" if you want to counter the input.

Dbase ID	Study
125024121	upregulated in Alzheimer's signature A
125025121	downregulated in Alzheimer's signature A
125026121	upregulated in Alzheimer's signature B1
125027121	downregulated in Alzheimer's signature B1
125028121	upregulated in Alzheimer's signature B2
125029121	downregulated in Alzheimer's signature B2
125030121	upregulated in Alzheimer's signature C1
125031121	downregulated in Alzheimer's signature C1
125032121	upregulated in Alzheimer's signature C2

As mentioned above, it's pretty obvious that there are major distinctions between the clusters. For example, the single best counter to Cluster A is Cluster C2. Unfortunately, negating one form of Alzheimer's with another is probably not a very good therapeutic option. More evidence of different forms of Alzheimer's is the fact that some individual Alzheimer's studies counter specific clusters. In the past, we've actually noted (and then ignored) this fact when comparing individual Alzheimer's studies.

There's a lot that could be said about the mimics and counters above. To keep things brief:

*not one but two studies involving alterations to ARX are seen above. Intriguing. 

*antiretrovirals are prominent but, as noted before, they were applied to HIV patients, meaning that we must choose between a "direct" effect on Alzheimer's or an indirect effect where the drugs kill a virus that induces an Alzheimer's-like transcriptome. Either way, though, the question of antiretroviral usage is interesting.

*there are a few easy interventions that are worth thinking about...e.g. DHA (fish oil) ingestion, zinc restriction, high protein diets. High activity levels counter the C clusters, but may enable the A cluster.

*some knockout/overexpression studies suggest possible drug targets besides ARX.

*mouse Alzheimer's models are nowhere to be seen in the mimics/counters columns. One mouse neuroinflammation study does appear, however, as mimicking cluster C1. If you're a fan of Alzheimer's mouse models, the paper does devote several paragraphs to this topic.

*cluster C1 is chock-full of neural disorders besides Alzheimer's. Why? Are we looking at a general response to insults to the brain?

*some results are slightly comical. It appears that alcoholism could counter the Cluster A version of Alzheimer's.

In general, the super-compelling clustering results suggest a future where treatment depends entirely on one's Alzheimer's type. Unfortunately, it's beyond the scope of the paper to show how patients might be typed, given the invasiveness of sampling living brain tissue. We note that our own "canonical" Alzheimer's lists strongly parallel clusters C1 and C2, probably because that's simply where the bulk of post-mortem samples abide. Also, the paper does show that tau tangles and AB-plaques are prominent in particular clusters; in-vivo imaging methods for these Alzheimer's manifestations are improving.

whatismygene.com 

Become a Patron!