Daniel Lemire's blog

, 4 min read

Science and Technology links (July 15th, 2018)

  1. One of the problems that occur with aging is that your immune system becomes less efficient, less able to learn.
    If we could reverse this effect, it would be akin to rejuvenation. And it seems to be around the corner according to a new article published by Science:

The objective of this phase 2a randomized, placebo-controlled clinical trial was to determine whether low-dose mTOR inhibitor therapy enhanced immune function and decreased infection rates in 264 elderly subjects given the study drugs for 6 weeks. A low-dose combination (…) was associated with a significant (P = 0.001) decrease in the rate of infections reported by elderly subjects for a year after study drug initiation. In addition, we observed an up-regulation of antiviral gene expression and an improvement in the response to influenza vaccination in this treatment group. Thus, selective TORC1 inhibition has the potential to improve immune function and reduce infections in the elderly.

  1. Nature reports on a technique that could massively accelerate artificial neural networks:

Our method replaces artificial neural networks fully-connected layers with sparse ones before training, reducing quadratically the number of parameters, with no decrease in accuracy.

Whether this makes it to production server is another story but “reducing quadratically the number of parameters” sounds impressive.

In related news, you can run neural network software on DNA to recognize molecular patterns.

  1. We live longer and longer, but what good is that if your mind is not intact? Thankfully, between 2000 and 2010, most of the increase in life expectancy has been concentrated in cognitively healthy years in this 10 year period.1. As we age, we accumulate “old” cells called senescent cells. They should die but they fail to do so. It seems that senescent cells are really bad for us. Thankfully we can clear them out and get benefits, the evidence is growing:

Here we demonstrate that transplanting relatively small numbers of senescent cells into young mice is sufficient to cause persistent physical dysfunction, as well as to spread cellular senescence to host tissues. Transplanting even fewer senescent cells had the same effect in older recipients and was accompanied by reduced survival, indicating the potency of senescent cells in shortening health- and lifespan. The senolytic cocktail, dasatinib plus quercetin, which causes selective elimination of senescent cells, decreased the number of naturally occurring senescent cells and their secretion of frailty-related proinflammatory cytokines in explants of human adipose tissue. Moreover, intermittent oral administration of senolytics to both senescent cell–transplanted young mice and naturally aged mice alleviated physical dysfunction and increased post-treatment survival by 36% while reducing mortality hazard to 65%.

  1. Vegetarian men are more likely to suffer from depression.
  2. Naked mole rats are ugly mammals that are interesting because they are largely immune to cancer and aging. Shar-Pei dogs are dogs with deeply wrinkled skin. Might they also be resistant to cancer? It looks like they are, indeed. However, sadly, Shar-Pei dogs are short lived due to other health problems.
  3. In England, more than half (57%) of those who die are older than 80.1. Australians show that you can wipe out disease-carrying mosquitoes using genetic engineering.
  4. A heart disease drug seems to be able to at least partially reverse Type 1 diabetes, in some cases:

addition of once-daily oral verapamil may be a safe and effective novel approach to promote endogenous beta cell function and reduce insulin requirements and hypoglycemic episodes in adult individuals with recent-onset T1D.

In related research, Omega-3 fatty acids could help with Type 1 diabetes.

  1. Though Siri and Alexa can understand what you are saying, most of the time, they cannot isolate your voice in a crowd of people talking. It is essentially an unsolved problem in speech recognition. It seems that some researchers made progress, bringing the error rate down from 90% to 30%, under some conditions. The problem remains unsolved (30% is high).