FINCHES-online

TL/DR

FINCHES-online lets you predict interactions between disordered regions using sequence as input.

You can put two sequences in to the Intermaps page, and it will generate a prediction of which subregions of each sequence are likely to be involved in intermolecular interactions.

You can also input two sequences into the Phase Diagrams page and predict two homotypic phase diagrams based on those sequences for easy comparison.

There are many caveats, and you should read about them if you plan to use FINCHES!

Introduction

Intrinsically disordered proteins and protein regions (IDRs) play crucial roles in various cellular processes. Unlike structured proteins, IDRs do not have a fixed three-dimensional structure. While IDRs can interact via folding-upon-binding, in many cases, their interaction with other biomolecules is facilitated and/or enhanced by chemical specificity, the favorable interaction between complementary chemical groups found in the disordered protein and its binding partner. These complementary interactions can enhance sequence-specific binding but may also drive intermolecular interactions in their own right. In the context of proteins that undergo phase separation in the test tube and form biomolecular condensates in cells, the underlying molecular grammar of IDR-mediated interactions is driven by this chemical specificity.

Structure-based tools such as AlphaFold offer the potential to predict IDR-mediated interactions where defined structural interfaces appear. However, these structure-centric methods are poorly suited for situations in which IDRs interact transiently with a partner with minimal (or no) acquisition of structure. As such, it generally remains challenging to determine which subregions of an IDR may be facilitating interactions with a partner. Importantly, this question is unavoidably partner-specific - the same IDR may interact with different partners in different ways, so there may not be a universal "binding region" within an IDR.

Here, we present a bottom-up approach that uses chemical physics extracted from coarse-grained force fields to predict IDR-mediated interactions. This approach is based on the premise that the chemical specificity of IDRs can be captured by the amino acid sequence alone (taking local sequence context into account) and that local attractive and repulsive interactions can be predicted and used to identify subregions within an IDR that have the potential to facilitate attractive or repulsive interactions. Our goal is to facilitate rapid and testable predictions for which residues and regions within an IDR are likely to be involved in interactions with a partner.

Application of FINCHES-online

Right now, FINCHES-online is suited only for interrogating IDR:IDR interactions. FINCHES does enable IDR:folded domain interactions to be investigated, and we will likely integrate this functionality into FINCHES-online in the coming months. However, IDR:IDR interactions are still likely useful for many applications, including the study of biomolecular condensates and other systems in which IDRs remain partially or fully disordered in their bound state.

High-level overview

Our method is based on the premise that the chemical specificity of IDRs can be captured by the amino acid sequence alone (taking local sequence context into account) and that local attractive and repulsive interactions can be predicted to identify subregions within an IDR that have the potential to facilitate attractive or repulsive interactions with a partner. Briefly, it does this by:

1. Calculating inter-residue interaction values based on the chemical physics from the force field being used
2. Modulate these naive calculations based on the local sequence context.
3. Either use this matrix to predict an overall mean-field interaction parameter which is used in phase diagram construction, or use a smoothing function to smooth across the sequence to identify subregions that mediate intermolecular attractive or repulsive interactions, as illustrated via a predicted intermolecular interaction map (intermap).

Caveats

This approach is fully analytical but has several core caveats:

1. Our method is sequence-based and does not consider the structural context of the IDR. As such, it is not well-suited for predicting IDR-mediated interactions if the IDR undergoes a large conformational change upon binding (e.g., folding).
2. Our method is not well-suited for predicting the quantitative energetics of IDR-mediated interactions. Instead, it is designed to identify regions within an IDR that are likely to be involved in interactions with a partner.
3. Our method is based on the premise that the chemical specificity of IDRs can be captured by the amino acid sequence alone and the local solution environment. However, all predictions on FINCHES-online necessitate a salt concentration of 150 mM NaCl under physiological buffer conditions. Tuning of the solution environment can be done using the local FINCHES package (to some extent).