1. Introduction

Analysis of gene concepts that capture past and present uses of the term has produced widespread skepticism about the prospects for uniting two key ways of characterizing genes—as causes of phenotypic effects and as genomic DNA sequences—in a coherent concept of the gene.Footnote ¹ The various ways DNA is involved in the processes that ultimately result in the most proximal effects of gene expression, the protein and RNA molecules often called “gene products,” seem to prohibit characterizing genes as entities with both a specific molecular product and a determinate physical identity as a segment of chromosomal DNA whose sequence corresponds to that of the gene product (Fogle Reference Griffiths, Stotz, Hull and Ruse2000; Moss Reference Moss2003; Downes Reference Eddy2004; Stotz, Bostanci, and Griffiths Reference Stotz, Bostanci and Griffiths2006). If there is no viable gene concept that combines both of these elements, then the research methods of genomics become somewhat mysterious. If analysis of a genome sequence cannot yield information about genes that are both defined in terms of their gene products and reflected in the DNA sequence—because there is no such thing—then what could we learn from it?Footnote ²

I will call the view that any gene concept incorporating both structural and functional specification of what a gene is cannot be coherently applied in genomic analysis the exclusivity thesis. In a recent article, Piotrowska applies the exclusivity thesis in her assessment of the results of comparative genomics. She starts by asking whether comparative genomics employs a gene concept that “maps onto a structural or a functional conception of the gene” (Reference Piotrowska2009, 245).Footnote ³ She claims that “[if] the gene employed as the unit of measurement corresponds to a structural gene concept, then comparative genomicists must view genes as ordered sequences of DNA,” and the comparative claims that result are merely claims about shared nucleic acid sequence (845). If, however, the gene concept involved is a functional concept, then “comparative genomicists must view genes as entities which give rise to phenotypes” and could produce substantive claims about similarity between the compared species (845). However, the analysis applied to genomic data involves searching sequences and evaluating sequences for similarity. Piotrowska's conclusion is that the results are not substantive; in other words, the comparisons are based on a structural gene concept rather than a functional one, so they do not have functional significance and so do not genuinely capture genetic similarity.

Piotrowska has made an important move in extending philosophical analysis of genes to the methods biologists use in research on genomes. Although philosophers of biology have made use of recent results from genomics, they have had surprisingly little to say about how genomes are analyzed, even in works organized around developments in genomics (Moss Reference Moss2006; Barnes and Dupré Reference Barnes and Dupré2008). Furthermore, while Piotrowska focuses on the most famous claims of comparative genomics, those reporting the percent of genes shared between two species, there is nothing in her analysis of the bioinformatic methods used in those studies that restricts her conclusion to these particular results. If Piotrowska is correct, it is unclear how the analytical tools applied to genomic sequences could yield any information of functional significance. But the sequence of a genome is of no use to biologists in the absence of the analytical tools used to study it; it is what biologists might learn from that sequence that matters. Given the resources, in funding and personnel, now devoted to sequencing and analyzing genomes, Piotrowska's paper raises a high-stakes question about the epistemology of genomic sequence analysis.Footnote ⁴

Since the problem of understanding what could be learned from analyzing genomic sequences depends on the exclusivity thesis, I will first review the biological phenomena that support the claim that genes cannot be entities specified both as genomic DNA segments and in terms of determinate molecular products. In the next section, I will explain some of the most basic techniques used in analyzing genomes. This will set the stage, first, for investigating whether the analytical techniques typically used in genomics depend on a particular gene concept and, second, to allow for clarification of the epistemic value of genomic research, given that this cannot depend on a gene concept that unifies structure and function.

2. The Gene Concept in Question

The philosophical literature on genes is too extensive to summarize here.Footnote ⁵ Suffice it to say that biologists have used the term ‘gene’ in so many different ways that, in response, philosophers have offered several different characterizations of types of gene concepts in order to make sense of the variety.Footnote ⁶ A survey of this variety is unnecessary for the purposes at hand since the epistemological questions raised about the study of genomic sequences arise from the two gene concepts that are most prominently involved in that research. I will first explain the biological phenomena that make it difficult to articulate a gene concept that combines these two—that could capture a physical specification of a gene as a nucleotide sequence and as something that is the cause of a functional molecular product. While the production of RNA molecules that play important functional roles in cellular processes is critical, in this article I will focus on the genes involved in protein production.

Two basic processes are necessary for protein production: transcription and translation. Transcription is the production of an RNA molecule from a segment of DNA, with the DNA providing a template for the sequence of nucleotides in the RNA. Translation is the production of an amino acid sequence from messenger RNA (mRNA); some transcribed RNA play different kinds of roles in the cell from mRNA, like ribosomal RNA (rRNA), and some RNA transcripts are called pre-mRNA because their sequences are altered to produce an mRNA that is then translated. In translation, the sequence of nucleotides in the mRNA directs the addition of amino acids, with three-nucleotide triples (codons) being used to determine which amino acid is added. The region of the mRNA that is ‘read’ in this way begins with a methionine codon that functions as a start signal, followed by the rest of the codons for the resulting amino acid chain, followed by a stop codon.

In prokaryotic organisms (bacteria and archaea), the RNA initially transcribed undergoes little processing before translation, and there is often a direct correspondence between the sequence of amino acids in the protein product and nucleotides in the open reading frame (ORF), the series of nucleotides from the start to the stop codon. However, this simple model is not an adequate account of gene expression in eukaryotes (animals, fungi, plants, and protists), which have significant post-transcriptional RNA processing. The result of those modifications is that the sequence of nucleotides in an ORF in the genome does not match the sequence of nucleotides in the mRNA that is involved in translation, and thus there is not a direct correspondence between the amino acid sequence of the protein product and the nucleotide sequence in the genome. Even in prokaryotes this correspondence may not hold, due to events like frameshifting, which occurs during the translation of some mRNAs: rather than ‘reading’ the nucleotide sequence as continuous, contiguous triples, the reading frame shifts one or more nucleotides during translation. In eukaryotes, the initial RNA transcript may go through extensive processing, typically cutting out some regions (introns) and splicing together others (exons—the parts of the original transcript that is expressed). Sometimes the RNA nucleotide sequence is simply modified, or edited, by changing a base from one nucleotide to another or adding or deleting bases. From the perspective of the resulting protein, there is no segment of genomic DNA with “physical integrity” (Fogle Reference Griffiths, Stotz, Hull and Ruse2000, 4) that corresponds to the amino acid sequence.

Furthermore, some initial RNA transcripts are spliced in alternative ways, depending on the circumstances, creating mRNAs with different subsets of the exons contained in the initial transcript. And there may be multiple initial transcripts produced from the same segment of the genome, by transcribing to capture different reading frames or by transcribing a segment with an ORF that is within a larger ORF. Thus, looking at things from the direction of a part of the genome, there is no one molecular product that can be associated with that segment as its genetic effect.

These implications have been stressed repeatedly by authors investigating the gene concept (Moss Reference Moss2003; Downes Reference Eddy2004; Stotz et al. Reference Stotz, Bostanci and Griffiths2006). Fogle claims that attempts to apply a consensus molecular gene that combines structural and functional features has the effect of “forcing arbitrary choices about the physical edges or the physiological properties of the gene” (Reference Griffiths, Stotz, Hull and Ruse2000, 19). This supports part of the assumption that Piotrowska makes: if genome analysis involves applying a gene concept in a nonarbitrary way, that concept involves either structural or functional specification, but not both. I will assume that these are indeed exclusive options. In the next section, I outline some basic analytical tools involved in research on genomes, setting the stage to assess whether that does indeed involve application of a gene concept and, if so, what sort of concept is involved.

3. Analyzing Genomes

There are multiple ways to analyze and compare genomes; in this section I focus on techniques that both are common and are likely candidates for involving gene concepts, in contrast to techniques like aligning whole genomes or determining and comparing guanine-cytosine content.

4. Genomic Analysis and the Gene Concept

The question raised by philosophical analysis of the gene concept is, if “genes” are not objects defined by both genomic sequence and molecular product, then how could analysis of genomic sequences produce results—purportedly about genes—that mean anything at all? Piotrowska's response to this problem is that comparative genomics cannot produce substantive—that is, functionally relevant—results. Her reasoning depends not only on the exclusivity thesis but also on an assumption of exhaustivity: she assumes that genomics must involve either a structural or a functional concept of the gene. Since the researchers in this area use the term “gene” frequently, this seems initially plausible. However, on closer investigation it is not clear that a gene concept is deeply involved in genomic research. The techniques used to analyze genomic sequences are not driven by a theory of the nature of the gene, and in particular, they are not driven by a structural concept of the gene. They draw on a variety of kinds of information about how DNA sequences are involved in the processes of transcription and translation and known statistical differences between the composition of coding and noncoding regions, background knowledge of the biochemistry involved in protein-DNA binding, and evolutionary theory. But use of this information does not imply that these analytical techniques are being used to investigate a kind of a thing that is defined as a segment of the genome. For example, aligning genomic sequences can be used to find orthologous segments of the genome, so these techniques make it possible to use a sequence from one organisms’ genome, known to be transcribed and spiced in alternate ways, to search for an ortholog in another organism. Because alignments can provide evidence of an evolutionary relationship between the two sequences, and that in turn provides some evidence that the two DNA sequences are involved in cellular processes, like transcription, in similar ways, this procedure offers a way to find other cases of alternative splicing.Footnote ⁸ These tools are not designed around a concept of the gene with “physical integrity.” The gene prediction tools may not be as successful in identifying disparate genomic DNA sequences that contribute to the production of a protein; they may be more prone to false negatives in such cases. The fact that the tools are less reliable in such cases does not show that their use involves a structural gene concept.

Similarly, the fact that the researchers are using these techniques to identify and compare putative molecular products does not imply that their results must be understood in terms of a functional gene concept, that is, finding and comparing genes defined as causes of molecular products. The analytical tools used in functional genomics can also be used to find and compare noncoding genomic sequences, which suggests the results of applying these tools should not be characterized as finding and comparing ‘genes’ defined in terms of molecular products.

If genomic research is not aimed at producing knowledge about the gene—defined either functionally or structurally—then what is its epistemic value? Genomic analysis allows for identifying causal relationships between parts of genomes and molecular products and identifying different kinds of evolutionary relationships among parts of genomes (such as orthologs and paralogs). These in turn have implications for whether the homologous genes share function. In general, the results require further support; these are discovery tools providing some evidence but not conclusive support. The epistemic importance of discovery in this context should not be underestimated: genome sequences can have billions of base pairs. There are hundreds of sequenced genomes, with more in the pipeline (see, e.g., the projects completed and in development at the Joint Genome Institute [Reference Downes2011]). It is only through the use of search tools that any hypotheses at all can be formulated from genomic sequence data.

5. Conclusion

This study is a preliminary effort to clarify the epistemology of genomic analysis. However, the survey of the most basic techniques involved has had several results. It shows that this kind of research is not constrained by a particular gene concept, nor are the results used to fill in or support a particular theory of the nature of the gene. Rather, various kinds of information about the relationships between genome sequences and protein production are used to find new cases. That information can involve stereotypical gene features, like the ORF structure, as well as information from specific cases (as when a known protein sequence is used as a query against a new genome). Higgs and Attwood characterize genomics as a “knowledge-based” discipline, to highlight the fact that it is not guided by some overarching theory but rather depends on many different kinds of facts known about organisms and evolution (2005). This study validates that characterization. Finally, identifying the reasoning involved in genome analysis as a discovery process both clarifies its epistemic value and sets the stage for further philosophical investigation.