Evaluating factors that influence microbial synthesis yields by linear regression with numerical and ordinal variables

Peter F. Colletti, Yogesh Goyal, Arul M. Varman, Xueyang Feng, Bing Wu, Yinjie J. Tang

Research output: Contribution to journalArticlepeer-review

22 Scopus citations


In the production of chemicals via microbial fermentation, achieving a high yield is one of the most important objectives. We developed a statistical model to analyze influential factors that determine product yield by compiling data obtained from engineered Escherichia coli developed within last 10 years. Using both numerical and ordinal variables (e.g., enzymatic steps, cultivation conditions, and genetic modifications) as input parameters, our model revealed that cultivation modes, nutrient supplementation, and oxygen conditions were the three significant factors for improving product yield. Generally, the model showed that product yield decreases as the number of enzymatic steps in the biosynthesis pathway increases (7-9% loss of yield per enzymatic step). Moreover, overexpression of enzymes or removal of competitive pathways (e.g., knockout) does not necessarily result in an amplification of product yield (P-value >0.1), possibly because of limited capacity in the biosynthesis pathway to accommodate an increase in flux. The model not only provides general guidelines for metabolic engineering and fermentation processes, but also allows a priori estimation and comparison of product yields under designed cultivation conditions.

Original languageEnglish (US)
Pages (from-to)893-901
Number of pages9
JournalBiotechnology and bioengineering
Issue number4
StatePublished - Apr 1 2011
Externally publishedYes


  • Enzymatic steps
  • Escherichia coli
  • Flux
  • Nutrients
  • Overexpression
  • P-value

ASJC Scopus subject areas

  • Biotechnology
  • Bioengineering
  • Applied Microbiology and Biotechnology


Dive into the research topics of 'Evaluating factors that influence microbial synthesis yields by linear regression with numerical and ordinal variables'. Together they form a unique fingerprint.

Cite this