What Google's Calico Anti-aging initiative will actually do | Hacker News Reader

What Google's Calico Anti-aging initiative will actually do

6 pointsdrcode3 comments

The biggest missing piece in curing most diseases IMHO is that we do not yet have any good computational model for a cell. If we had a way to simulate an individual cell for a specific person in a computer program (which is a hard problem) we could actually cure cancers pretty easily (by testing millions of pharmaceutical compounds on the spot against each person's cancer cell model) and also make great strides into all other human diseases.

Only a very few scientists have worked in the field of whole cell simulation so far (see http://goo.gl/heurd for one effort) and this problem is a PERFECT application for Google's big data expertise.

Here's what I'm guessing Calico will do:

1. Fund scientists who work on technologies that can detect various chemical concentrations of cells in vivo.

2. Build a multi-petabyte database of chemical concentration timelines generated in this way, which are run against human cell cultures experiencing varying environmental conditions.

3. Run complex big data statistical algorithms to determine quantitative measures of causation between all these chemicals and with DNA sequences in the cell.

4. Use this information to build an accurate computational model of a human cell.

This is an area of research that (1) would improve therapies across the board for all major diseases at once (2) has few short-term financial incentives, meaning only a huge research initiative could attempt it, and (3) requires significant big data expertise.

Thoughts?

Mapping, modeling, and indexing a cell's structure is a very different problem than indexing hyperlink structure on the internet.

A wholesale computational model of the cell is an incredibly ambitious task to achieve via a broad, unfocused approach to building a chem-concentration database.

The main issue they would face is in your bullet #1: What is your "Pagerank" for cell structure and chem-concentration analysis, what/who do you put your weight behind?

How do you merge these data sets? This research is distributed in labs all over the planet with different quality standards, research objectives, lab conditions, data hygiene, statistical significance, etc.

What you've written seems like a plausible approach in a pure research setting within a single lab, but if their end goal is a computational model of the cell with commercial side-effects along the way-- the best way to achieve that would be to chip away at it piecemeal. Not through an unfocused merger of cell "big data".

E.g. identify specific protein binding sites and conformations that correlate with a specific type of breast cancer, track their conformations under different conditions--- prove that your data is superior to data yielded by existing models.

Of course, you would also have to monetize somehow.

The specific thing you chose is less important than not attempting to paint the whole model at once.

Although something akin to this should occur, I think your post understates the computational enormity of the issue. Many diseases are not caused by single genes or single environmental conditions, but interactions between dozens of genes and the environment, even epigenetic factors. Big Data (huge data) approaches will begin to chip away at the mountain, but even with Google's super powers, it's a ways off.

> (1) would improve therapies across the board for all major diseases at once

How? Unless if by "at once" you mean they would begin the era of using this approach on any disease. But surely each disease will have its own puzzle.

Re: Computationally enormity:

Well, I agree the computational enormity is a big problem, but certainly some biochemical pathways will be more amenable to this type of approach than others, and so it should be possible to make incremental progress by solving one puzzle at a time.

Re: Using on any disease

I agree having this sort of technology is not an immediate solution to each arbitrary disease. However, being able to accurately predict how different cells in a specific person's body behave to different drug cocktails would give doctors "super powers" that are almost unimaginable today. A doctor with a patient that has cholesterol plaques might say "If we give him these ten drugs in this dosage his T lymphocytes will aggressively attack the plaques in the LAD artery only, and the LAD arterial cells will increase division to reduce risk of subsequent aneurysm from autoimmune side-effects".

If the accuracy of the computational model is good enough, it might be just as effective as the old "nanobots in the bloodstream" concept, but much easier to implement. In effect, the patient's own cells could function as the "nanobots" by reprogramming them

Reply on news.ycombinator.com