Sigmoid

What it is

In the systems-biology source, Sigmoid is a generative, scalable software infrastructure built to support modeling of biological pathways. It helps biologists and computational scientists create and explore predictive dynamical models of complex systems such as metabolic, gene regulation, and signal transduction pathways. In the machine-learning source, sigmoid means a smooth S-shaped function; the specific formula shown is the logistic function, which maps any real input to a value between 0 and 1.

[1] A Scalable and Integrative System for Pathway Bioinformatics and Systems Biology [2] Neural networks: Activation functions | Machine Learning

How the software Sigmoid works

The software is organized as distributed modules in a three-tier architecture: a database, a simulator and other model manipulators in the back end, and a graphical user interface in the front end connected through web-service middleware. Its core modules include pathway or cell model generation and simulation, a pathway modeling database, a web service-oriented middleware layer, a web model browser, and a biologist-friendly graphical interface. A key goal is to translate biological knowledge into mathematical form, with symbolic reactions converted into mathematical equations and solved by the system.

[1] A Scalable and Integrative System for Pathway Bioinformatics and Systems Biology

How the mathematical sigmoid works

The sigmoid function takes an input x and transforms it with the formula 1/(1+e^-x), where e is the base of natural logarithms. The output is always between 0 and 1, so it can be used as a probability-like score. In a neural network, an activation function is the rule that transforms a node’s output before it is passed to the next layer; sigmoid is one example, alongside tanh and ReLU.

[2] Neural networks: Activation functions | Machine Learning [3] 25 logistic regression - Stanford University

Limitations and cautions

The term is ambiguous: one source uses Sigmoid as software infrastructure, while the other sources use sigmoid for a mathematical function. For the function, the sources note that other activation functions such as ReLU can be easier to compute and may be less affected by the vanishing gradient problem during training. For the software, the source describes a research infrastructure and does not claim it is a complete or universal model of biology.

[1] A Scalable and Integrative System for Pathway Bioinformatics and Systems Biology [2] Neural networks: Activation functions | Machine Learning

Why the software matters

Sigmoid was built to support a cycle of model building, hypothesis generation, biological experimentation, and data gathering. It also connects to external data sources and analysis tools through web services, and the described version had been populated with more than 20 published models ranging from simple molecular interactions to cell-fate decision networks.

[1] A Scalable and Integrative System for Pathway Bioinformatics and Systems Biology

Sources and further reading

This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.

1. A Scalable and Integrative System for Pathway Bioinformatics and Systems Biology
2. Neural networks: Activation functions | Machine Learning
3. 25 logistic regression - Stanford University