**7. Conclusions**

In this review we have explored the ways in which bioreceptor-free biosensors can benefit from ML methods. Robust ML models bring specificity and accuracy to array-based biosensors such as Enose and Etongue by learning the patterns in the sensor responses. Notably, PCA has shown grea<sup>t</sup> performance as a feature extraction technique for these systems. Similar power of PCA has been demonstrated for optical biosensors that generate spectra such as Raman spectra or SERS. ANNs using deep learning generate impressive results for imaging-based sensors including lensless holography and digital staining. ML has also been used in creative ways such as for data fusion of multiple biosensors, and transfer learning for noise correction, sensor drift compensation, and domain adaptation.

However, many practical challenges still exist. Many of the methods presented here are not widely used in commercial settings. This is due to many reasons including variability in manufacturing and the ability to make compact versions of the biosensors while maintaining performance. ML models that can adapt to differences in sensor response are at an advantage, and transfer learning shows promise to be part of the solution.

In recent years, ML has garnered strong research interest in many fields including biosensing, as evidenced in this review. If this review has inspired interest to learn more about how machine learning is being used for one of the methods presented here, we encourage you to seek more specific reviews for the subject. There are grea<sup>t</sup> reviews in the literature, many of which were referenced, that take a closer look at the methods presented in this review.

**Author Contributions:** Conceptualization, K.E.S.III and J.-Y.Y.; Methodology, K.E.S.III; Formal analysis, K.E.S.III; Investigation, K.E.S.III; Data curation, K.E.S.III; Writing—original draft preparation, K.E.S.III; Writing—review and editing, J.-Y.Y.; Supervision, J.-Y.Y.; Project administration, J.-Y.Y.; Funding acquisition, K.E.S.III and J.-Y.Y. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the U.S. National Institutes of Health under the gran<sup>t</sup> T32GM132008.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable. **Data Availability Statement:** This study did not report any data.

**Acknowledgments:** The authors would like to thank Lane E. Breshears for her contribution to the collection of papers and for discussions on how to organize this manuscript. The authors would also like to thank Kattika Kaarj (now at Mahidol University) for helpful discussions and proofreading this manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.
