bioinformatics

syllabus

general information

homework

lectures

websites

Will Terzaghi's Homepage

Membership

Login

Week 14 lecture

• Two general approaches have been taken to try to estimate the minimal number of genes needed by a cell.

1. Researchers at TIGR and the University of North Carolina performed saturation mutagenesis on Mycoplasma genitalium, the bacterium with the smallest known genome (517 genes).
   • they classified every mutation that gave a lethal phenotype as a necessary gene
   • using this approach they estimate that Mycoplasma genitalium needs between 265 and 350 genes
   • However. note
     • they didn't actually trim the genome down to 350 genes: they mutated one gene at a time and counted up the number that gave a lethal phenotype
     • Although Mycoplasma genitalium is free-living, it is a pathogen that must take up many complex nutrients from its host
2. The alternative is to compare genomes of many different organisms and identify those that all had in common: the theoretical minimum genome.
   • Comparison of the genomes of Mycoplasma genitalium and Haemophilus influenzae identified 250 genes in common (diverged ~ 1.5 billion years ago)
   • However, both are human parasites

   Note that both approaches come up with about 250 genes as the estimate of the bare minimum for a free-living cell.

   Dr. Pier Luisi in Zurich, Switzerland is taking a very different approach: his group is trying to build cells from scratch!

   • One way is to construct liposomes with proteins and/or RNA molecules trapped inside
   • Another is to create "reverse micelles" with macromoleucles trapped inside
   • A final way is to construct giant vesicles and then inject proteins and/or RNA molecules

1. E-cell is one example
   • the original version constructed a model using 127 genes from M. genitalium sufficient for transcription, translation, energy production and phospholipid synthesis.
   • difference with programs such as GEPASI was that it attempted to integrate several different pathways
   • was able to simulate life, but not replicate it.
   • one problem, don't fully understand how mycoplasmas divide.
2. Are now up to the third generation of E-cell
   • E-cell 2 is designed to run on windows, and can be downloaded from http://www.bioinformatics.org/%7Enaota/ecell2/
   • E-cell3 is runs on UNIX
   • Are now modeling E-coli, erythrocytes and rice.
3. Gene Network sciences is a company which is using whole cell simulations for drug discovery http://www.gnsbiotech.com/minimalcell.shtml

Artificial life

Workers in this field are trying to answer two questions

1. can non-organic life-forms multiply and evolve?
2. can non-organic life-forms become sentient?

   Basically, they are trying to study the physics of the living state.

1) can non-organic life-forms multiply and evolve?

• Artificial life is defined as non-biological systems which evolve to greater levels of fitness by natural selection.
  • algorithms must meet a fitness test before they reproduce
  • fitness of each subsequent generation improves according to the selective criteria
• Rationale: by synthesizing life in alternative media, we can distinguish essential properties of life from properties that are found in all terrestrial life-forms due to common genetic descent.
  • Simple living systems are populations of self-replicating entities that harbor information in the form of coded symbols
  • if we assume basic principles governing living systems are independent of the substrate, then life can be recreated in a computational medium
    • special-purpose computer programs replicating in core memory have a lot in common with bacteria replicating in a petri dish!

This is a very serious and active area of research! eg http://dllab.caltech.edu/research/

It also has its own organization http://www.alife.org/ and its own journal published by MIT press! http://mitpress.mit.edu/catalog/item/default.asp?ttype=4&;tid=41

Lots of programs have been developed pioneered by Ray with the tierra software http://www.isd.atr.co.jp/~ray/tierra/ Links to many are posted at http://staff.aist.go.jp/utsugi-a/Lab/Links.html

• Some, such as SimEarth or SimLife are just games,
• Others, such as Avida, are serious research tools! http://dllab.caltech.edu/avida/

A real-world application of alife: A drug company in Southern California uses an evolutionary program to find the best means of docking an inhibitory drug molecule to a protein essential to the reproduction of HIV

2) can non-organic life-forms become sentient?

Can we teach computers to think?

• This is the ultimate goal of artificial intelligence research!
• Turing proposed the following test in 1950: if a computer can repeatedly fool a human interrogator in a different room that a human is answering his/her questions, then the computer is intelligent.
• Basic problem is teaching computers to model human psychology! This has proven to be extremely difficult!
• Example: language relies heavily on context, so designing translation programs has been extremely difficult. Humans still do it better!
  • a classic example: after translating from English to Russian then back to English "The spirit is willing but the flesh is weak" became "The wine is agreeable but the meat is rotten."
• Many people have tried to write programs for writing fiction
  • A simple (but fun) example is the complaint letter generator http://hugin.sigusr1.org/~pakin/complaint
  • A more serious example is Erasmotron: models human psychology in order to teach computers to write fiction http://www.erasmatazz.com/userdoc.html
• Other programs have been written to try to hold conversations with humans.
  • ELIZA was one of the first: it simulated a psychoanalyst holding a conversation with a patient, with bizarre results
  • PC-therapist was more successful, partly because it randomly introduced typos!
• General approach has now shifted to solving subsets of artificial intelligence piece by piece, rather than all at once.
  • We've already seen that neural networks are very good at pattern recognition: something that humans traditionally do much better than computers
  • Another approach used is "machine learning." This is used by game programs, where if a particular move pays off, they are more likely to use it in subsequent games.
  • Expert systems are designed to solve specific problems, such as diagnosis, using "fuzzy logic" based on probabilities.
    • Grammar checkers are an example of an expert system

• However, people are still trying! e.g. Ramona, the first live virtual performing and recording artist. http://www.kurzweilai.net/meme/frame.html?m=9

Self-replicating machines

First proposed in 1940s by John von Neumann

His machine consisted of two separate components:

1. A universal computer capable of computing anything it was instructed to
2. A universal constructor that can make anything universal computer tells it to

Self-replication involved two different processes:

1. generating copies of the computer and of the constructor
2. making a copy of the program and attaching it to the computer

Cellular automata are the computational model which he developed for self-replicating machines

• an array of cells, each of which can be in one of a finite number of possible states, updated according to a local interaction rule.
• The state of a cell at the next time step is determined by the current states of surrounding cells.
• used two-dimensional CAs with 29 states per cell and a neighborhood consisting of 5 cells
• described how to build a universal constructor : a machine capable of constructing any configuration whose description can be stored on its “input tape”
• His basic design has since been confirmed to work
• His universal constructor has been the inspiration for subsequent attempts at self-replicating machines.
• Moshe Sipper provides an annotated bibliography of articles on self-replicating machines at http://www.cs.bgu.ac.il/~sipper/selfrep/

One of the most thorough studies of self-replicating machines was conducted by NASA in 1980 is discuss the feasibility of using self-replicating machines to colonize the moon and planets (posted at http://www.islandone.org/MMSG/aasm/ )

• Rationale:gravity limits the quantity of material that can be exported into space
• Much better to export minimal devices that can replicate themselves ( as well as make other things) using locally available materials
• Don’t need much information!
• Need even less if broadcast instructions (Also prevents them from becoming sentient)
• Plan was to send 100 tons of seed material into space
• Factories would contain mining robots etc capable of making items including new factories
• Advantages : no requirement for human workers and little strain on earth’s resources

Currently there is a lot of speculation about the possibility of constructing self-replicating nanomachines ( as you will have seen while reading Bill Joy's concerns)

However, the only place I found that actually claims to have constructed a self-replicating machine is the GOLEM project at Brandeis http://demo.cs.brandeis.edu/golem/index.html

They developed robots which could design and build other robots!