If one is given the complete genome of some unknown organism, would it be possible to systematically deduce what this organism looks like and behaves like without reference to anything else (ex: a repository of genomes of known living organisms)?

In another words, do we currently have a model capable of translating any random genome of an organism to a graphical representation of the organism and information on its behaviour?

Are there such 'genome-structure' or 'genome-behaviour' relationships known to us that would enable us to accomplish such a task?


As you specifically state:

Would it be possible to systematically deduce what this organism looks like and behaves like without reference to anything else (ex: a repository of genomes of known living organisms)?

No, no chance at all, we certainly wouldn't even be able to determine a single gene function, other than that it codes for a series of amino-acids. After all, even simple genes of say 100 bases in DNA will still have 33 amino-acids, and most people don't remember that sort of information without cause, and the only way we can acquire that easily information is through repositories of information. The only way we could do something approaching working this out would be to take portions of the genome that we could identify as genes (start/stop codons), then express them one-by-one and empirically work out the function through classic biochemical methods. This is very laborious and time consuming - it would take many many person-hours per gene and a LOT of resources. However, this is how functions of novel genes and their related proteins were and still are worked out.

With reference - perhaps, at least partially. From the sequence and with access to a repository such as Genbank we could deduce which type of organism it was by comparing to known genomes from other organisms in a process known as phylogenetic analysis. You could certainly tell which of the kingdoms it came from. In each organism there are conserved bits of the DNA (or RNA) that tell you which groups they are. For example all DNA based bacteria and archaea (that I know of) you can use the 16S ribosomal RNA sequence, which will identify bacteria and archaea down to genus and species level.

For instance, you might have a genome that gives you information that it is a bacterium of the Staphylcoccus group, meaning that you can deduce that it is very very likely that it will be a small bacterium which is a round ball shape, and will stain positive with a Gram stain. You could also likely identify some genes and their function based on homology.

You can apply this process to other genomes, for instance a mammalian one will tell you that it had fur, produced milk, was warm blooded, quadruped, spinal column etc.


No, current technology is nowhere near what is required to deduce the form of a species from its DNA alone without comparing it to the DNA of the same or similar species.

The reason for this is that genes in the DNA generate proteins and these proteins are simple building blocks that combine in highly complex ways that we do not understand. Currently scientists are simply trying to learn how to predict the SHAPE of the protein from its DNA sequence. That is the first step. We are getting close to being able to do that. Once we can predict shape, then we can start working on figuring how that shape translates into function.

I would also add that organisms have many expressed traits. That means that the trait expresses itself only after the organism grows and is not genetically determined. For example, fingerprints are an expressed trait. Also, colorings and pigmentation are expressed traits. For example, I have two cats that are genetic twins and have exactly the same same DNA, but one has a small patch of white fur on his chest and the other does not.

  • $\begingroup$ However, there are individual genetic variants with known function. For example, given a person's genome, we should be able to roughly predict what hair color and texture they have, eye color, etc. $\endgroup$
    – Armand
    Aug 10 at 2:05
  • 2
    $\begingroup$ @Armand, that doesn't fit the question, which specifies "without reference to anything else (ex: a repository of genomes of known living organisms)". $\endgroup$
    – mgkrebbs
    Aug 10 at 7:49
  • $\begingroup$ @Armand we know only that specific genes correlate with specific hair colors, but we don't know the full cause-and-effect chain. In the end, to determine "this gene encodes red hair" from first principles, you'd have to know the function of all the genes that are involved in creating hair cells, pigments in general, and how the gene is activated in hair cells but not in skin cells. $\endgroup$
    – toolforger
    Aug 10 at 13:27
  • 1
    $\begingroup$ Yes, by OPs mentioning a "model" I was assuming that model would contain info like hair color genes and so forth. In re-reading the question it does seem the OP intends a de novo model based on structure prediction. $\endgroup$
    – Armand
    Aug 10 at 13:31

Echoing other answers, our ability to predict the function of individual genes is entirely dependent on either A) physical experiments (making the products of those genes in a test tube or performing genetic experiments) or B) inferring their function from their similarity to other genes that have been studied by physical experiments. In other words, we can't currently predict individual gene functions with computers unless we have some similar outside reference to compare to. Without the ability to do that for a single gene, there is no possibility of doing this for a whole genome. Forget about predicting how all of these genes would express or interact to create a complex living organism. If you could compare to other sequenced genomes, phylogenetic analysis would quickly tell you what species your unknown genome was most similar to.

I thought I might offer an alternate answer about what one might be able to do with some general knowledge of genomes but not the ability to actually compare your unknown genome to specific genomes / known DNA sequence. With the right software, you could roughly predict where the genes were in your unknown genome. This could tell you how many genes this organism has, how many of those genes are unique genes or repeats, and how much of the genome is made of non-genic sequence. You could also look at the overall structure of your genome. These factors could give you a rough guess about what kind of organism you have. For example, if your genome is circular (which you could tell from the sequence), you have a prokaryote or archaeon. If there are very few genes, you could infer that this organism lived a parasitic lifestyle, since the smallest known genomes are from parasitic species that no longer need to encode for all of the functions of a fully independent lifestyle. A large genome, linear chromosomes, or multiple chromosomes would indicate a eukaryote. If you see evidence of polyploidy, your organism could be a plant. Through similar methods, you could probably tell a eukaryote's nuclear genome apart from mitochondrial and chloroplast genomes and guess whether it was a plant based on how many non-nuclear chromosomes it has (two distinct non-nuclear chromosomes would indicate a chloroplast genome is present). Just a little thought experiment on what one could do with an unknown genome sequence without Genbank or reference sequence, but with current computational tools and some knowledge about genomes.

  • $\begingroup$ How would you know (without reference to genbank or similar) that a particular gene came from a mitochondria or a chloroplast? I can see the bit about eukaryote vs prokaryote, but I don't see that the next steps would work. I don't think people have solved how to determine ploidy with sequencing yet - please let me know if I am wrong, because I love to learn new methods. $\endgroup$
    – bob1
    Aug 10 at 20:57
  • $\begingroup$ Ploidy is a number-of-copies measurement. If you had error-free sequence for all the DNA from a single cell, multiple alleles for the same gene could be tallied separately. You might reasonably infer triploidy if you found many genes with three alleles present, tetraploidy if you found many with four alleles present, etc. Of course you wouldn't expect multiple alleles at every site, but finding even a few such would be a strong hint. $\endgroup$
    – Ethan
    Aug 10 at 21:34
  • $\begingroup$ @bob1, for mitochondria or chloroplast chromosomes, they're both circular (although wikipedia tells me that there's some debate about chloroplasts). So, they should be identifiable apart from other eukaryotic chromosomes without having to worry about identifying the genes themselves. Also, I'm guessing that both mitochondria and chloroplast chromosomes are much more gene-rich than most nuclear chromosomes, so they could probably be confirmed with that as well. $\endgroup$
    – LTRretro
    Aug 11 at 14:51
  • $\begingroup$ @bob1, as to polyploidy, I had allopolyploids in mind like wheat and other plants, which I think would be strikingly obvious if you were to look carefully at your sequence. I'm not experienced in this area at all, but Ethans answer sounds pretty logical to me for cases like autopolyploidy. That would depend on having a fair amount of heterozygous alleles though. $\endgroup$
    – LTRretro
    Aug 11 at 15:02
  • $\begingroup$ @Ethan - but modern sequencing technology (with perhaps the exception of PacBio and Oxford Nanopore) gives varying sequencing coverage and depth for a single sample, which in no way relates to ploidy as far as I can tell. There are tools out there to work this sort of thing out, but as far as I can tell they are definitely imperfect to put it mildly. $\endgroup$
    – bob1
    Aug 11 at 20:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.